X-Git-Url: https://wimlib.net/git/?p=wimlib;a=blobdiff_plain;f=src%2Flzx-compress.c;h=a972cc35bf3c24722372db9538c3b9a72f21f1c8;hp=ae1cf152a51da785085b16813c0a5b372c84076e;hb=7953f731d41d728a8881872bcf82fd8f9d1f7ee8;hpb=35ff480326a41f7f74fe0a497c636a811922f276 diff --git a/src/lzx-compress.c b/src/lzx-compress.c index ae1cf152..a972cc35 100644 --- a/src/lzx-compress.c +++ b/src/lzx-compress.c @@ -1,11 +1,11 @@ /* * lzx-compress.c * - * LZX compression routines + * A compressor that produces output compatible with the LZX compression format. */ /* - * Copyright (C) 2012, 2013 Eric Biggers + * Copyright (C) 2012, 2013, 2014 Eric Biggers * * This file is part of wimlib, a library for working with WIM files. * @@ -25,515 +25,377 @@ /* - * This file contains a compressor for the LZX compression format, as used in - * the WIM file format. + * This file contains a compressor for the LZX ("Lempel-Ziv eXtended") + * compression format, as used in the WIM (Windows IMaging) file format. * - * Format - * ====== + * Two different parsing algorithms are implemented: "near-optimal" and "lazy". + * "Near-optimal" is significantly slower than "lazy", but results in a better + * compression ratio. The "near-optimal" algorithm is used at the default + * compression level. * - * First, the primary reference for the LZX compression format is the - * specification released by Microsoft. + * This file may need some slight modifications to be used outside of the WIM + * format. In particular, in other situations the LZX block header might be + * slightly different, and a sliding window rather than a fixed-size window + * might be required. * - * Second, the comments in lzx-decompress.c provide some more information about - * the LZX compression format, including errors in the Microsoft specification. - * - * Do note that LZX shares many similarities with DEFLATE, the algorithm used by - * zlib and gzip. Both LZX and DEFLATE use LZ77 matching and Huffman coding, - * and certain other details are quite similar, such as the method for storing - * Huffman codes. However, some of the main differences are: + * Note: LZX is a compression format derived from DEFLATE, the format used by + * zlib and gzip. Both LZX and DEFLATE use LZ77 matching and Huffman coding. + * Certain details are quite similar, such as the method for storing Huffman + * codes. However, the main differences are: * * - LZX preprocesses the data to attempt to make x86 machine code slightly more * compressible before attempting to compress it further. + * * - LZX uses a "main" alphabet which combines literals and matches, with the * match symbols containing a "length header" (giving all or part of the match - * length) and a "position slot" (giving, roughly speaking, the order of + * length) and an "offset slot" (giving, roughly speaking, the order of * magnitude of the match offset). - * - LZX does not have static Huffman blocks; however it does have two types of - * dynamic Huffman blocks ("aligned offset" and "verbatim"). - * - LZX has a minimum match length of 2 rather than 3. - * - In LZX, match offsets 0 through 2 actually represent entries in an LRU - * queue of match offsets. This is very useful for certain types of files, - * such as binary files that have repeating records. - * - * Algorithms - * ========== - * - * There are actually two distinct overall algorithms implemented here. We - * shall refer to them as the "slow" algorithm and the "fast" algorithm. The - * "slow" algorithm spends more time compressing to achieve a higher compression - * ratio compared to the "fast" algorithm. More details are presented below. - * - * Slow algorithm - * -------------- - * - * The "slow" algorithm to generate LZX-compressed data is roughly as follows: - * - * 1. Preprocess the input data to translate the targets of x86 call - * instructions to absolute offsets. - * - * 2. Build the suffix array and inverse suffix array for the input data. The - * suffix array contains the indices of all suffixes of the input data, - * sorted lexcographically by the corresponding suffixes. The "position" of - * a suffix is the index of that suffix in the original string, whereas the - * "rank" of a suffix is the index at which that suffix's position is found - * in the suffix array. - * - * 3. Build the longest common prefix array corresponding to the suffix array. - * - * 4. For each suffix, find the highest lower ranked suffix that has a lower - * position, the lowest higher ranked suffix that has a lower position, and - * the length of the common prefix shared between each. This information is - * later used to link suffix ranks into a doubly-linked list for searching - * the suffix array. - * - * 5. Set a default cost model for matches/literals. - * - * 6. Determine the lowest cost sequence of LZ77 matches ((offset, length) - * pairs) and literal bytes to divide the input into. Raw match-finding is - * done by searching the suffix array using a linked list to avoid - * considering any suffixes that start after the current position. Each run - * of the match-finder returns the approximate lowest-cost longest match as - * well as any shorter matches that have even lower approximate costs. Each - * such run also adds the suffix rank of the current position into the linked - * list being used to search the suffix array. Parsing, or match-choosing, - * is solved as a minimum-cost path problem using a forward "optimal parsing" - * algorithm based on the Deflate encoder from 7-Zip. This algorithm moves - * forward calculating the minimum cost to reach each byte until either a - * very long match is found or until a position is found at which no matches - * start or overlap. - * - * 7. Build the Huffman codes needed to output the matches/literals. - * - * 8. Up to a certain number of iterations, use the resulting Huffman codes to - * refine a cost model and go back to Step #6 to determine an improved - * sequence of matches and literals. - * - * 9. Output the resulting block using the match/literal sequences and the - * Huffman codes that were computed for the block. - * - * Note: the algorithm does not yet attempt to split the input into multiple LZX - * blocks, instead using a series of blocks of LZX_DIV_BLOCK_SIZE bytes. * - * Fast algorithm - * -------------- + * - LZX does not have static Huffman blocks (that is, the kind with preset + * Huffman codes); however it does have two types of dynamic Huffman blocks + * ("verbatim" and "aligned"). * - * The fast algorithm (and the only one available in wimlib v1.5.1 and earlier) - * spends much less time on the main bottlenecks of the compression process --- - * that is, the match finding and match choosing. Matches are found and chosen - * with hash chains using a greedy parse with one position of look-ahead. No - * block splitting is done; only compressing the full input into an aligned - * offset block is considered. + * - LZX has a minimum match length of 2 rather than 3. Length 2 matches can be + * useful, but generally only if the parser is smart about choosing them. * - * API - * === - * - * The old API (retained for backward compatibility) consists of just one - * function: - * - * wimlib_lzx_compress() - * - * The new compressor has more potential parameters and needs more memory, so - * the new API ties up memory allocations and compression parameters into a - * context: - * - * wimlib_lzx_alloc_context() - * wimlib_lzx_compress2() - * wimlib_lzx_free_context() - * wimlib_lzx_set_default_params() - * - * Both wimlib_lzx_compress() and wimlib_lzx_compress2() are designed to - * compress an in-memory buffer of up to the window size, which can be any power - * of two between 2^15 and 2^21 inclusively. However, by default, the WIM - * format uses 2^15, and this is seemingly the only value that is compatible - * with WIMGAPI. In any case, the window is not a true "sliding window" since - * no data is ever "slid out" of the window. This is needed for the WIM format, - * which is designed such that chunks may be randomly accessed. - * - * Both wimlib_lzx_compress() and wimlib_lzx_compress2() return 0 if the data - * could not be compressed to less than the size of the uncompressed data. - * Again, this is suitable for the WIM format, which stores such data chunks - * uncompressed. - * - * The functions in this LZX compression API are exported from the library, - * although with the possible exception of wimlib_lzx_set_default_params(), this - * is only in case other programs happen to have uses for it other than WIM - * reading/writing as already handled through the rest of the library. - * - * Acknowledgments - * =============== - * - * Acknowledgments to several open-source projects and research papers that made - * it possible to implement this code: - * - * - divsufsort (author: Yuta Mori), for the suffix array construction code, - * located in a separate directory (divsufsort/). - * - * - "Linear-Time Longest-Common-Prefix Computation in Suffix Arrays and Its - * Applications" (Kasai et al. 2001), for the LCP array computation. - * - * - "LPF computation revisited" (Crochemore et al. 2009) for the prev and next - * array computations. - * - * - 7-Zip (author: Igor Pavlov) for the algorithm for forward optimal parsing - * (match-choosing). - * - * - zlib (author: Jean-loup Gailly and Mark Adler), for the hash table - * match-finding algorithm (used in lz77.c). - * - * - lzx-compress (author: Matthew T. Russotto), on which some parts of this - * code were originally based. + * - In LZX, offset slots 0 through 2 actually represent entries in an LRU queue + * of match offsets. This is very useful for certain types of files, such as + * binary files that have repeating records. */ #ifdef HAVE_CONFIG_H # include "config.h" #endif -#include "wimlib.h" -#include "wimlib/compress.h" +#include "wimlib/compress_common.h" +#include "wimlib/compressor_ops.h" #include "wimlib/endianness.h" #include "wimlib/error.h" +#include "wimlib/lz_mf.h" +#include "wimlib/lz_repsearch.h" #include "wimlib/lzx.h" #include "wimlib/util.h" -#include -#include + #include +#include -#ifdef ENABLE_LZX_DEBUG -# include "wimlib/decompress.h" -#endif +#define LZX_OPTIM_ARRAY_LENGTH 4096 -#include "divsufsort/divsufsort.h" +#define LZX_DIV_BLOCK_SIZE 32768 -typedef u32 block_cost_t; -#define INFINITE_BLOCK_COST ((block_cost_t)~0U) +#define LZX_CACHE_PER_POS 8 -#define LZX_OPTIM_ARRAY_SIZE 4096 +#define LZX_MAX_MATCHES_PER_POS (LZX_MAX_MATCH_LEN - LZX_MIN_MATCH_LEN + 1) -#define LZX_DIV_BLOCK_SIZE 32768 +#define LZX_CACHE_LEN (LZX_DIV_BLOCK_SIZE * (LZX_CACHE_PER_POS + 1)) -#define LZX_MAX_CACHE_PER_POS 10 +struct lzx_compressor; -/* Codewords for the LZX main, length, and aligned offset Huffman codes */ +/* Codewords for the LZX Huffman codes. */ struct lzx_codewords { - u16 main[LZX_MAINCODE_MAX_NUM_SYMBOLS]; - u16 len[LZX_LENCODE_NUM_SYMBOLS]; - u16 aligned[LZX_ALIGNEDCODE_NUM_SYMBOLS]; + u32 main[LZX_MAINCODE_MAX_NUM_SYMBOLS]; + u32 len[LZX_LENCODE_NUM_SYMBOLS]; + u32 aligned[LZX_ALIGNEDCODE_NUM_SYMBOLS]; }; -/* Codeword lengths (in bits) for the LZX main, length, and aligned offset - * Huffman codes. - * - * A 0 length means the codeword has zero frequency. - */ +/* Codeword lengths (in bits) for the LZX Huffman codes. + * A zero length means the corresponding codeword has zero frequency. */ struct lzx_lens { u8 main[LZX_MAINCODE_MAX_NUM_SYMBOLS]; u8 len[LZX_LENCODE_NUM_SYMBOLS]; u8 aligned[LZX_ALIGNEDCODE_NUM_SYMBOLS]; }; -/* Costs for the LZX main, length, and aligned offset Huffman symbols. - * - * If a codeword has zero frequency, it must still be assigned some nonzero cost - * --- generally a high cost, since even if it gets used in the next iteration, - * it probably will not be used very times. */ +/* Estimated cost, in bits, to output each symbol in the LZX Huffman codes. */ struct lzx_costs { u8 main[LZX_MAINCODE_MAX_NUM_SYMBOLS]; u8 len[LZX_LENCODE_NUM_SYMBOLS]; u8 aligned[LZX_ALIGNEDCODE_NUM_SYMBOLS]; }; -/* The LZX main, length, and aligned offset Huffman codes */ +/* Codewords and lengths for the LZX Huffman codes. */ struct lzx_codes { struct lzx_codewords codewords; struct lzx_lens lens; }; -/* Tables for tallying symbol frequencies in the three LZX alphabets */ +/* Symbol frequency counters for the LZX Huffman codes. */ struct lzx_freqs { - input_idx_t main[LZX_MAINCODE_MAX_NUM_SYMBOLS]; - input_idx_t len[LZX_LENCODE_NUM_SYMBOLS]; - input_idx_t aligned[LZX_ALIGNEDCODE_NUM_SYMBOLS]; + u32 main[LZX_MAINCODE_MAX_NUM_SYMBOLS]; + u32 len[LZX_LENCODE_NUM_SYMBOLS]; + u32 aligned[LZX_ALIGNEDCODE_NUM_SYMBOLS]; }; -/* LZX intermediate match/literal format */ -struct lzx_match { - /* Bit Description - * - * 31 1 if a match, 0 if a literal. - * - * 30-25 position slot. This can be at most 50, so it will fit in 6 - * bits. - * - * 8-24 position footer. This is the offset of the real formatted - * offset from the position base. This can be at most 17 bits - * (since lzx_extra_bits[LZX_MAX_POSITION_SLOTS - 1] is 17). - * - * 0-7 length of match, minus 2. This can be at most - * (LZX_MAX_MATCH_LEN - 2) == 255, so it will fit in 8 bits. */ - u32 data; -}; +/* Intermediate LZX match/literal format */ +struct lzx_item { -/* Raw LZ match/literal format: just a length and offset. - * - * The length is the number of bytes of the match, and the offset is the number - * of bytes back in the input the match is from the current position. - * - * If @len < LZX_MIN_MATCH_LEN, then it's really just a literal byte and @offset is - * meaningless. */ -struct raw_match { - u16 len; - input_idx_t offset; + /* Bits 0 - 9: Main symbol + * Bits 10 - 17: Length symbol + * Bits 18 - 22: Number of extra offset bits + * Bits 23+ : Extra offset bits */ + u64 data; }; -/* Specification for an LZX block. */ -struct lzx_block_spec { - - /* One of the LZX_BLOCKTYPE_* constants indicating which type of this - * block. */ - int block_type; - - /* 0-based position in the window at which this block starts. */ - input_idx_t window_pos; - - /* The number of bytes of uncompressed data this block represents. */ - input_idx_t block_size; - - /* The position in the 'chosen_matches' array in the `struct - * lzx_compressor' at which the match/literal specifications for - * this block begin. */ - input_idx_t chosen_matches_start_pos; - - /* The number of match/literal specifications for this block. */ - input_idx_t num_chosen_matches; - - /* Huffman codes for this block. */ - struct lzx_codes codes; +/* Internal compression parameters */ +struct lzx_compressor_params { + u32 (*choose_items_for_block)(struct lzx_compressor *, u32, u32); + u32 num_optim_passes; + enum lz_mf_algo mf_algo; + u32 min_match_length; + u32 nice_match_length; + u32 max_search_depth; }; /* - * An array of these structures is used during the match-choosing algorithm. - * They correspond to consecutive positions in the window and are used to keep - * track of the cost to reach each position, and the match/literal choices that - * need to be chosen to reach that position. + * Match chooser position data: + * + * An array of these structures is used during the near-optimal match-choosing + * algorithm. They correspond to consecutive positions in the window and are + * used to keep track of the cost to reach each position, and the match/literal + * choices that need to be chosen to reach that position. */ -struct lzx_optimal { - /* The approximate minimum cost, in bits, to reach this position in the - * window which has been found so far. */ - block_cost_t cost; - - /* The union here is just for clarity, since the fields are used in two - * slightly different ways. Initially, the @prev structure is filled in - * first, and links go from later in the window to earlier in the - * window. Later, @next structure is filled in and links go from - * earlier in the window to later in the window. */ - union { - struct { - /* Position of the start of the match or literal that - * was taken to get to this position in the approximate - * minimum-cost parse. */ - input_idx_t link; - - /* Offset (as in an LZ (length, offset) pair) of the - * match or literal that was taken to get to this - * position in the approximate minimum-cost parse. */ - input_idx_t match_offset; - } prev; - struct { - /* Position at which the match or literal starting at - * this position ends in the minimum-cost parse. */ - input_idx_t link; - - /* Offset (as in an LZ (length, offset) pair) of the - * match or literal starting at this position in the - * approximate minimum-cost parse. */ - input_idx_t match_offset; - } next; - }; - - /* The match offset LRU queue that will exist when the approximate - * minimum-cost path to reach this position is taken. */ - struct lzx_lru_queue queue; -}; +struct lzx_mc_pos_data { -/* Suffix array link */ -struct salink { - /* Rank of highest ranked suffix that has rank lower than the suffix - * corresponding to this structure and either has a lower position - * (initially) or has a position lower than the highest position at - * which matches have been searched for so far, or -1 if there is no - * such suffix. */ - input_idx_t prev; - - /* Rank of lowest ranked suffix that has rank greater than the suffix - * corresponding to this structure and either has a lower position - * (intially) or has a position lower than the highest position at which - * matches have been searched for so far, or -1 if there is no such - * suffix. */ - input_idx_t next; - - /* Length of longest common prefix between the suffix corresponding to - * this structure and the suffix with rank @prev, or 0 if @prev is -1. - */ - input_idx_t lcpprev; + /* The cost, in bits, of the lowest-cost path that has been found to + * reach this position. This can change as progressively lower cost + * paths are found to reach this position. */ + u32 cost; +#define MC_INFINITE_COST UINT32_MAX - /* Length of longest common prefix between the suffix corresponding to - * this structure and the suffix with rank @next, or 0 if @next is -1. + /* The match or literal that was taken to reach this position. This can + * change as progressively lower cost paths are found to reach this + * position. + * + * This variable is divided into two bitfields. + * + * Literals: + * Low bits are 1, high bits are the literal. + * + * Explicit offset matches: + * Low bits are the match length, high bits are the offset plus 2. + * + * Repeat offset matches: + * Low bits are the match length, high bits are the queue index. */ - input_idx_t lcpnext; -}; + u32 mc_item_data; +#define MC_OFFSET_SHIFT 9 +#define MC_LEN_MASK ((1 << MC_OFFSET_SHIFT) - 1) -/* State of the LZX compressor. */ + /* The state of the LZX recent match offsets queue at this position. + * This is filled in lazily, only after the minimum-cost path to this + * position is found. + * + * Note: the way we handle this adaptive state in the "minimum-cost" + * parse is actually only an approximation. It's possible for the + * globally optimal, minimum cost path to contain a prefix, ending at a + * position, where that path prefix is *not* the minimum cost path to + * that position. This can happen if such a path prefix results in a + * different adaptive state which results in lower costs later. We do + * not solve this problem; we only consider the lowest cost to reach + * each position, which seems to be an acceptable approximation. */ + struct lzx_lru_queue queue _aligned_attribute(16); + +} _aligned_attribute(16); + +/* State of the LZX compressor */ struct lzx_compressor { - /* The parameters that were used to create the compressor. */ - struct wimlib_lzx_params params; + /* Internal compression parameters */ + struct lzx_compressor_params params; - /* The buffer of data to be compressed. - * - * 0xe8 byte preprocessing is done directly on the data here before - * further compression. - * - * Note that this compressor does *not* use a real sliding window!!!! - * It's not needed in the WIM format, since every chunk is compressed - * independently. This is by design, to allow random access to the - * chunks. - * - * We reserve a few extra bytes to potentially allow reading off the end - * of the array in the match-finding code for optimization purposes. - */ - u8 *window; + /* The preprocessed buffer of data being compressed */ + u8 *cur_window; /* Number of bytes of data to be compressed, which is the number of - * bytes of data in @window that are actually valid. */ - input_idx_t window_size; + * bytes of data in @cur_window that are actually valid. */ + u32 cur_window_size; - /* Allocated size of the @window. */ - input_idx_t max_window_size; - - /* Number of symbols in the main alphabet (depends on the - * @max_window_size since it determines the maximum allowed offset). */ + /* log2 order of the LZX window size for LZ match offset encoding + * purposes. Will be >= LZX_MIN_WINDOW_ORDER and <= + * LZX_MAX_WINDOW_ORDER. + * + * Note: 1 << @window_order is normally equal to @max_window_size, + * a.k.a. the allocated size of @cur_window, but it will be greater than + * @max_window_size in the event that the compressor was created with a + * non-power-of-2 block size. (See lzx_get_window_order().) */ + unsigned window_order; + + /* Number of symbols in the main alphabet. This depends on + * @window_order, since @window_order determines the maximum possible + * offset. It does not, however, depend on the *actual* size of the + * current data buffer being processed, which might be less than 1 << + * @window_order. */ unsigned num_main_syms; + /* Lempel-Ziv match-finder */ + struct lz_mf *mf; + + /* Match-finder wrapper functions and data for near-optimal parsing. + * + * When doing more than one match-choosing pass over the data, matches + * found by the match-finder are cached to achieve a slight speedup when + * the same matches are needed on subsequent passes. This is suboptimal + * because different matches may be preferred with different cost + * models, but it is a very worthwhile speedup. */ + unsigned (*get_matches_func)(struct lzx_compressor *, const struct lz_match **); + void (*skip_bytes_func)(struct lzx_compressor *, unsigned n); + u32 match_window_pos; + u32 match_window_end; + struct lz_match *cached_matches; + struct lz_match *cache_ptr; + struct lz_match *cache_limit; + + /* Position data for near-optimal parsing. */ + struct lzx_mc_pos_data optimum[LZX_OPTIM_ARRAY_LENGTH + LZX_MAX_MATCH_LEN]; + + /* The cost model currently being used for near-optimal parsing. */ + struct lzx_costs costs; + /* The current match offset LRU queue. */ struct lzx_lru_queue queue; - /* Space for the sequences of matches/literals that were chosen for each - * block. */ - struct lzx_match *chosen_matches; + /* Frequency counters for the current block. */ + struct lzx_freqs freqs; - /* Information about the LZX blocks the preprocessed input was divided - * into. */ - struct lzx_block_spec *block_specs; + /* The Huffman codes for the current and previous blocks. */ + struct lzx_codes codes[2]; - /* Number of LZX blocks the input was divided into; a.k.a. the number of - * elements of @block_specs that are valid. */ - unsigned num_blocks; + /* Which 'struct lzx_codes' is being used for the current block. The + * other was used for the previous block (if this isn't the first + * block). */ + unsigned int codes_index; - /* This is simply filled in with zeroes and used to avoid special-casing - * the output of the first compressed Huffman code, which conceptually - * has a delta taken from a code with all symbols having zero-length - * codewords. */ - struct lzx_codes zero_codes; + /* Dummy lengths that are always 0. */ + struct lzx_lens zero_lens; - /* The current cost model. */ - struct lzx_costs costs; + /* Matches/literals that were chosen for the current block. */ + struct lzx_item chosen_items[LZX_DIV_BLOCK_SIZE]; - /* Fast algorithm only: Array of hash table links. */ - input_idx_t *prev_tab; + /* Table mapping match offset => offset slot for small offsets */ +#define LZX_NUM_FAST_OFFSETS 32768 + u8 offset_slot_fast[LZX_NUM_FAST_OFFSETS]; +}; - /* Suffix array for window. - * This is a mapping from suffix rank to suffix position. */ - input_idx_t *SA; +/* + * Structure to keep track of the current state of sending bits to the + * compressed output buffer. + * + * The LZX bitstream is encoded as a sequence of 16-bit coding units. + */ +struct lzx_output_bitstream { - /* Inverse suffix array for window. - * This is a mapping from suffix position to suffix rank. - * If 0 <= r < window_size, then ISA[SA[r]] == r. */ - input_idx_t *ISA; + /* Bits that haven't yet been written to the output buffer. */ + u32 bitbuf; - /* Longest common prefix array corresponding to the suffix array SA. - * LCP[i] is the length of the longest common prefix between the - * suffixes with positions SA[i - 1] and SA[i]. LCP[0] is undefined. - */ - input_idx_t *LCP; + /* Number of bits currently held in @bitbuf. */ + u32 bitcount; - /* Suffix array links. - * - * During a linear scan of the input string to find matches, this array - * used to keep track of which rank suffixes in the suffix array appear - * before the current position. Instead of searching in the original - * suffix array, scans for matches at a given position traverse a linked - * list containing only suffixes that appear before that position. */ - struct salink *salink; - - /* Position in window of next match to return. */ - input_idx_t match_window_pos; - - /* The match-finder shall ensure the length of matches does not exceed - * this position in the input. */ - input_idx_t match_window_end; - - /* Matches found by the match-finder are cached in the following array - * to achieve a slight speedup when the same matches are needed on - * subsequent passes. This is suboptimal because different matches may - * be preferred with different cost models, but seems to be a worthwhile - * speedup. */ - struct raw_match *cached_matches; - unsigned cached_matches_pos; - bool matches_cached; - - /* Slow algorithm only: Temporary space used for match-choosing - * algorithm. - * - * The size of this array must be at least LZX_MAX_MATCH_LEN but - * otherwise is arbitrary. More space simply allows the match-choosing - * algorithm to potentially find better matches (depending on the input, - * as always). */ - struct lzx_optimal *optimum; + /* Pointer to the start of the output buffer. */ + le16 *start; - /* Slow algorithm only: Variables used by the match-choosing algorithm. - * - * When matches have been chosen, optimum_cur_idx is set to the position - * in the window of the next match/literal to return and optimum_end_idx - * is set to the position in the window at the end of the last - * match/literal to return. */ - u32 optimum_cur_idx; - u32 optimum_end_idx; + /* Pointer to the position in the output buffer at which the next coding + * unit should be written. */ + le16 *next; + + /* Pointer past the end of the output buffer. */ + le16 *end; }; -/* Returns the LZX position slot that corresponds to a given match offset, - * taking into account the recent offset queue and updating it if the offset is - * found in it. */ -static unsigned -lzx_get_position_slot(unsigned offset, struct lzx_lru_queue *queue) +/* + * Initialize the output bitstream. + * + * @os + * The output bitstream structure to initialize. + * @buffer + * The buffer being written to. + * @size + * Size of @buffer, in bytes. + */ +static void +lzx_init_output(struct lzx_output_bitstream *os, void *buffer, u32 size) +{ + os->bitbuf = 0; + os->bitcount = 0; + os->start = buffer; + os->next = os->start; + os->end = os->start + size / sizeof(le16); +} + +/* + * Write some bits to the output bitstream. + * + * The bits are given by the low-order @num_bits bits of @bits. Higher-order + * bits in @bits cannot be set. At most 17 bits can be written at once. + * + * @max_num_bits is a compile-time constant that specifies the maximum number of + * bits that can ever be written at the call site. Currently, it is used to + * optimize away the conditional code for writing a second 16-bit coding unit + * when writing fewer than 17 bits. + * + * If the output buffer space is exhausted, then the bits will be ignored, and + * lzx_flush_output() will return 0 when it gets called. + */ +static inline void +lzx_write_varbits(struct lzx_output_bitstream *os, + const u32 bits, const unsigned int num_bits, + const unsigned int max_num_bits) { - unsigned position_slot; - - /* See if the offset was recently used. */ - for (unsigned i = 0; i < LZX_NUM_RECENT_OFFSETS; i++) { - if (offset == queue->R[i]) { - /* Found it. */ - - /* Bring the repeat offset to the front of the - * queue. Note: this is, in fact, not a real - * LRU queue because repeat matches are simply - * swapped to the front. */ - swap(queue->R[0], queue->R[i]); - - /* The resulting position slot is simply the first index - * at which the offset was found in the queue. */ - return i; + /* This code is optimized for LZX, which never needs to write more than + * 17 bits at once. */ + LZX_ASSERT(num_bits <= 17); + LZX_ASSERT(num_bits <= max_num_bits); + LZX_ASSERT(os->bitcount <= 15); + + /* Add the bits to the bit buffer variable. @bitcount will be at most + * 15, so there will be just enough space for the maximum possible + * @num_bits of 17. */ + os->bitcount += num_bits; + os->bitbuf = (os->bitbuf << num_bits) | bits; + + /* Check whether any coding units need to be written. */ + if (os->bitcount >= 16) { + + os->bitcount -= 16; + + /* Write a coding unit, unless it would overflow the buffer. */ + if (os->next != os->end) + *os->next++ = cpu_to_le16(os->bitbuf >> os->bitcount); + + /* If writing 17 bits, a second coding unit might need to be + * written. But because 'max_num_bits' is a compile-time + * constant, the compiler will optimize away this code at most + * call sites. */ + if (max_num_bits == 17 && os->bitcount == 16) { + if (os->next != os->end) + *os->next++ = cpu_to_le16(os->bitbuf); + os->bitcount = 0; } } +} - /* The offset was not recently used; look up its real position slot. */ - position_slot = lzx_get_position_slot_raw(offset + LZX_OFFSET_OFFSET); +/* Use when @num_bits is a compile-time constant. Otherwise use + * lzx_write_varbits(). */ +static inline void +lzx_write_bits(struct lzx_output_bitstream *os, + const u32 bits, const unsigned int num_bits) +{ + lzx_write_varbits(os, bits, num_bits, num_bits); +} + +/* + * Flush the last coding unit to the output buffer if needed. Return the total + * number of bytes written to the output buffer, or 0 if an overflow occurred. + */ +static u32 +lzx_flush_output(struct lzx_output_bitstream *os) +{ + if (os->next == os->end) + return 0; - /* Bring the new offset to the front of the queue. */ - for (unsigned i = LZX_NUM_RECENT_OFFSETS - 1; i > 0; i--) - queue->R[i] = queue->R[i - 1]; - queue->R[0] = offset; + if (os->bitcount != 0) + *os->next++ = cpu_to_le16(os->bitbuf << (16 - os->bitcount)); - return position_slot; + return (const u8 *)os->next - (const u8 *)os->start; } /* Build the main, length, and aligned offset Huffman codes used in LZX. @@ -541,8 +403,7 @@ lzx_get_position_slot(unsigned offset, struct lzx_lru_queue *queue) * This takes as input the frequency tables for each code and produces as output * a set of tables that map symbols to codewords and codeword lengths. */ static void -lzx_make_huffman_codes(const struct lzx_freqs *freqs, - struct lzx_codes *codes, +lzx_make_huffman_codes(const struct lzx_freqs *freqs, struct lzx_codes *codes, unsigned num_main_syms) { make_canonical_huffman_code(num_main_syms, @@ -564,405 +425,283 @@ lzx_make_huffman_codes(const struct lzx_freqs *freqs, codes->codewords.aligned); } -/* - * Output an LZX match. - * - * @out: The bitstream to write the match to. - * @block_type: The type of the LZX block (LZX_BLOCKTYPE_ALIGNED or LZX_BLOCKTYPE_VERBATIM) - * @match: The match. - * @codes: Pointer to a structure that contains the codewords for the - * main, length, and aligned offset Huffman codes. - */ -static void -lzx_write_match(struct output_bitstream *out, int block_type, - struct lzx_match match, const struct lzx_codes *codes) +static unsigned +lzx_compute_precode_items(const u8 lens[restrict], + const u8 prev_lens[restrict], + const unsigned num_lens, + u32 precode_freqs[restrict], + unsigned precode_items[restrict]) { - /* low 8 bits are the match length minus 2 */ - unsigned match_len_minus_2 = match.data & 0xff; - /* Next 17 bits are the position footer */ - unsigned position_footer = (match.data >> 8) & 0x1ffff; /* 17 bits */ - /* Next 6 bits are the position slot. */ - unsigned position_slot = (match.data >> 25) & 0x3f; /* 6 bits */ - unsigned len_header; - unsigned len_footer; - unsigned main_symbol; - unsigned num_extra_bits; - unsigned verbatim_bits; - unsigned aligned_bits; - - /* If the match length is less than MIN_MATCH_LEN (= 2) + - * NUM_PRIMARY_LENS (= 7), the length header contains - * the match length minus MIN_MATCH_LEN, and there is no - * length footer. - * - * Otherwise, the length header contains - * NUM_PRIMARY_LENS, and the length footer contains - * the match length minus NUM_PRIMARY_LENS minus - * MIN_MATCH_LEN. */ - if (match_len_minus_2 < LZX_NUM_PRIMARY_LENS) { - len_header = match_len_minus_2; - /* No length footer-- mark it with a special - * value. */ - len_footer = (unsigned)(-1); - } else { - len_header = LZX_NUM_PRIMARY_LENS; - len_footer = match_len_minus_2 - LZX_NUM_PRIMARY_LENS; - } - - /* Combine the position slot with the length header into a single symbol - * that will be encoded with the main code. - * - * The actual main symbol is offset by LZX_NUM_CHARS because values - * under LZX_NUM_CHARS are used to indicate a literal byte rather than a - * match. */ - main_symbol = ((position_slot << 3) | len_header) + LZX_NUM_CHARS; - - /* Output main symbol. */ - bitstream_put_bits(out, codes->codewords.main[main_symbol], - codes->lens.main[main_symbol]); - - /* If there is a length footer, output it using the - * length Huffman code. */ - if (len_footer != (unsigned)(-1)) { - bitstream_put_bits(out, codes->codewords.len[len_footer], - codes->lens.len[len_footer]); - } - - num_extra_bits = lzx_get_num_extra_bits(position_slot); - - /* For aligned offset blocks with at least 3 extra bits, output the - * verbatim bits literally, then the aligned bits encoded using the - * aligned offset code. Otherwise, only the verbatim bits need to be - * output. */ - if ((block_type == LZX_BLOCKTYPE_ALIGNED) && (num_extra_bits >= 3)) { + unsigned *itemptr; + unsigned run_start; + unsigned run_end; + unsigned extra_bits; + int delta; + u8 len; + + itemptr = precode_items; + run_start = 0; + do { + /* Find the next run of codeword lengths. */ - verbatim_bits = position_footer >> 3; - bitstream_put_bits(out, verbatim_bits, - num_extra_bits - 3); + /* len = the length being repeated */ + len = lens[run_start]; - aligned_bits = (position_footer & 7); - bitstream_put_bits(out, - codes->codewords.aligned[aligned_bits], - codes->lens.aligned[aligned_bits]); - } else { - /* verbatim bits is the same as the position - * footer, in this case. */ - bitstream_put_bits(out, position_footer, num_extra_bits); - } -} + run_end = run_start + 1; -static unsigned -lzx_build_precode(const u8 lens[restrict], - const u8 prev_lens[restrict], - const unsigned num_syms, - input_idx_t precode_freqs[restrict LZX_PRECODE_NUM_SYMBOLS], - u8 output_syms[restrict num_syms], - u8 precode_lens[restrict LZX_PRECODE_NUM_SYMBOLS], - u16 precode_codewords[restrict LZX_PRECODE_NUM_SYMBOLS], - unsigned *num_additional_bits_ret) -{ - memset(precode_freqs, 0, - LZX_PRECODE_NUM_SYMBOLS * sizeof(precode_freqs[0])); - - /* Since the code word lengths use a form of RLE encoding, the goal here - * is to find each run of identical lengths when going through them in - * symbol order (including runs of length 1). For each run, as many - * lengths are encoded using RLE as possible, and the rest are output - * literally. - * - * output_syms[] will be filled in with the length symbols that will be - * output, including RLE codes, not yet encoded using the precode. - * - * cur_run_len keeps track of how many code word lengths are in the - * current run of identical lengths. */ - unsigned output_syms_idx = 0; - unsigned cur_run_len = 1; - unsigned num_additional_bits = 0; - for (unsigned i = 1; i <= num_syms; i++) { - - if (i != num_syms && lens[i] == lens[i - 1]) { - /* Still in a run--- keep going. */ - cur_run_len++; + /* Fast case for a single length. */ + if (likely(run_end == num_lens || len != lens[run_end])) { + delta = prev_lens[run_start] - len; + if (delta < 0) + delta += 17; + precode_freqs[delta]++; + *itemptr++ = delta; + run_start++; continue; } - /* Run ended! Check if it is a run of zeroes or a run of - * nonzeroes. */ - - /* The symbol that was repeated in the run--- not to be confused - * with the length *of* the run (cur_run_len) */ - unsigned len_in_run = lens[i - 1]; + /* Extend the run. */ + do { + run_end++; + } while (run_end != num_lens && len == lens[run_end]); - if (len_in_run == 0) { - /* A run of 0's. Encode it in as few length - * codes as we can. */ + if (len == 0) { + /* Run of zeroes. */ - /* The magic length 18 indicates a run of 20 + n zeroes, - * where n is an uncompressed literal 5-bit integer that - * follows the magic length. */ - while (cur_run_len >= 20) { - unsigned additional_bits; - - additional_bits = min(cur_run_len - 20, 0x1f); - num_additional_bits += 5; + /* Symbol 18: RLE 20 to 51 zeroes at a time. */ + while ((run_end - run_start) >= 20) { + extra_bits = min((run_end - run_start) - 20, 0x1f); precode_freqs[18]++; - output_syms[output_syms_idx++] = 18; - output_syms[output_syms_idx++] = additional_bits; - cur_run_len -= 20 + additional_bits; + *itemptr++ = 18 | (extra_bits << 5); + run_start += 20 + extra_bits; } - /* The magic length 17 indicates a run of 4 + n zeroes, - * where n is an uncompressed literal 4-bit integer that - * follows the magic length. */ - while (cur_run_len >= 4) { - unsigned additional_bits; - - additional_bits = min(cur_run_len - 4, 0xf); - num_additional_bits += 4; + /* Symbol 17: RLE 4 to 19 zeroes at a time. */ + if ((run_end - run_start) >= 4) { + extra_bits = min((run_end - run_start) - 4, 0xf); precode_freqs[17]++; - output_syms[output_syms_idx++] = 17; - output_syms[output_syms_idx++] = additional_bits; - cur_run_len -= 4 + additional_bits; + *itemptr++ = 17 | (extra_bits << 5); + run_start += 4 + extra_bits; } - } else { /* A run of nonzero lengths. */ - /* The magic length 19 indicates a run of 4 + n - * nonzeroes, where n is a literal bit that follows the - * magic length, and where the value of the lengths in - * the run is given by an extra length symbol, encoded - * with the precode, that follows the literal bit. - * - * The extra length symbol is encoded as a difference - * from the length of the codeword for the first symbol - * in the run in the previous code. - * */ - while (cur_run_len >= 4) { - unsigned additional_bits; - signed char delta; - - additional_bits = (cur_run_len > 4); - num_additional_bits += 1; - delta = (signed char)prev_lens[i - cur_run_len] - - (signed char)len_in_run; + /* Symbol 19: RLE 4 to 5 of any length at a time. */ + while ((run_end - run_start) >= 4) { + extra_bits = (run_end - run_start) > 4; + delta = prev_lens[run_start] - len; if (delta < 0) delta += 17; precode_freqs[19]++; - precode_freqs[(unsigned char)delta]++; - output_syms[output_syms_idx++] = 19; - output_syms[output_syms_idx++] = additional_bits; - output_syms[output_syms_idx++] = delta; - cur_run_len -= 4 + additional_bits; + precode_freqs[delta]++; + *itemptr++ = 19 | (extra_bits << 5) | (delta << 6); + run_start += 4 + extra_bits; } } - /* Any remaining lengths in the run are outputted without RLE, - * as a difference from the length of that codeword in the - * previous code. */ - while (cur_run_len > 0) { - signed char delta; - - delta = (signed char)prev_lens[i - cur_run_len] - - (signed char)len_in_run; + /* Output any remaining lengths without RLE. */ + while (run_start != run_end) { + delta = prev_lens[run_start] - len; if (delta < 0) delta += 17; - - precode_freqs[(unsigned char)delta]++; - output_syms[output_syms_idx++] = delta; - cur_run_len--; + precode_freqs[delta]++; + *itemptr++ = delta; + run_start++; } + } while (run_start != num_lens); - cur_run_len = 1; - } - - /* Build the precode from the frequencies of the length symbols. */ - - make_canonical_huffman_code(LZX_PRECODE_NUM_SYMBOLS, - LZX_MAX_PRE_CODEWORD_LEN, - precode_freqs, precode_lens, - precode_codewords); - - *num_additional_bits_ret = num_additional_bits; - - return output_syms_idx; + return itemptr - precode_items; } /* - * Writes a compressed Huffman code to the output, preceded by the precode for - * it. - * - * The Huffman code is represented in the output as a series of path lengths - * from which the canonical Huffman code can be reconstructed. The path lengths - * themselves are compressed using a separate Huffman code, the precode, which - * consists of LZX_PRECODE_NUM_SYMBOLS (= 20) symbols that cover all possible - * code lengths, plus extra codes for repeated lengths. The path lengths of the - * precode precede the path lengths of the larger code and are uncompressed, - * consisting of 20 entries of 4 bits each. - * - * @out: Bitstream to write the code to. - * @lens: The code lengths for the Huffman code, indexed by symbol. - * @prev_lens: Code lengths for this Huffman code, indexed by symbol, - * in the *previous block*, or all zeroes if this is the - * first block. - * @num_syms: The number of symbols in the code. + * Output a Huffman code in the compressed form used in LZX. + * + * The Huffman code is represented in the output as a logical series of codeword + * lengths from which the Huffman code, which must be in canonical form, can be + * reconstructed. + * + * The codeword lengths are themselves compressed using a separate Huffman code, + * the "precode", which contains a symbol for each possible codeword length in + * the larger code as well as several special symbols to represent repeated + * codeword lengths (a form of run-length encoding). The precode is itself + * constructed in canonical form, and its codeword lengths are represented + * literally in 20 4-bit fields that immediately precede the compressed codeword + * lengths of the larger code. + * + * Furthermore, the codeword lengths of the larger code are actually represented + * as deltas from the codeword lengths of the corresponding code in the previous + * block. + * + * @os: + * Bitstream to which to write the compressed Huffman code. + * @lens: + * The codeword lengths, indexed by symbol, in the Huffman code. + * @prev_lens: + * The codeword lengths, indexed by symbol, in the corresponding Huffman + * code in the previous block, or all zeroes if this is the first block. + * @num_lens: + * The number of symbols in the Huffman code. */ static void -lzx_write_compressed_code(struct output_bitstream *out, +lzx_write_compressed_code(struct lzx_output_bitstream *os, const u8 lens[restrict], const u8 prev_lens[restrict], - unsigned num_syms) + unsigned num_lens) { - input_idx_t precode_freqs[LZX_PRECODE_NUM_SYMBOLS]; - u8 output_syms[num_syms]; + u32 precode_freqs[LZX_PRECODE_NUM_SYMBOLS]; u8 precode_lens[LZX_PRECODE_NUM_SYMBOLS]; - u16 precode_codewords[LZX_PRECODE_NUM_SYMBOLS]; + u32 precode_codewords[LZX_PRECODE_NUM_SYMBOLS]; + unsigned precode_items[num_lens]; + unsigned num_precode_items; + unsigned precode_item; + unsigned precode_sym; unsigned i; - unsigned num_output_syms; - u8 precode_sym; - unsigned dummy; - - num_output_syms = lzx_build_precode(lens, - prev_lens, - num_syms, - precode_freqs, - output_syms, - precode_lens, - precode_codewords, - &dummy); - - /* Write the lengths of the precode codes to the output. */ + for (i = 0; i < LZX_PRECODE_NUM_SYMBOLS; i++) - bitstream_put_bits(out, precode_lens[i], - LZX_PRECODE_ELEMENT_SIZE); + precode_freqs[i] = 0; - /* Write the length symbols, encoded with the precode, to the output. */ + /* Compute the "items" (RLE / literal tokens and extra bits) with which + * the codeword lengths in the larger code will be output. */ + num_precode_items = lzx_compute_precode_items(lens, + prev_lens, + num_lens, + precode_freqs, + precode_items); - for (i = 0; i < num_output_syms; ) { - precode_sym = output_syms[i++]; + /* Build the precode. */ + make_canonical_huffman_code(LZX_PRECODE_NUM_SYMBOLS, + LZX_MAX_PRE_CODEWORD_LEN, + precode_freqs, precode_lens, + precode_codewords); - bitstream_put_bits(out, precode_codewords[precode_sym], - precode_lens[precode_sym]); - switch (precode_sym) { - case 17: - bitstream_put_bits(out, output_syms[i++], 4); - break; - case 18: - bitstream_put_bits(out, output_syms[i++], 5); - break; - case 19: - bitstream_put_bits(out, output_syms[i++], 1); - bitstream_put_bits(out, - precode_codewords[output_syms[i]], - precode_lens[output_syms[i]]); - i++; - break; - default: - break; + /* Output the lengths of the codewords in the precode. */ + for (i = 0; i < LZX_PRECODE_NUM_SYMBOLS; i++) + lzx_write_bits(os, precode_lens[i], LZX_PRECODE_ELEMENT_SIZE); + + /* Output the encoded lengths of the codewords in the larger code. */ + for (i = 0; i < num_precode_items; i++) { + precode_item = precode_items[i]; + precode_sym = precode_item & 0x1F; + lzx_write_varbits(os, precode_codewords[precode_sym], + precode_lens[precode_sym], + LZX_MAX_PRE_CODEWORD_LEN); + if (precode_sym >= 17) { + if (precode_sym == 17) { + lzx_write_bits(os, precode_item >> 5, 4); + } else if (precode_sym == 18) { + lzx_write_bits(os, precode_item >> 5, 5); + } else { + lzx_write_bits(os, (precode_item >> 5) & 1, 1); + precode_sym = precode_item >> 6; + lzx_write_varbits(os, precode_codewords[precode_sym], + precode_lens[precode_sym], + LZX_MAX_PRE_CODEWORD_LEN); + } } } } +/* Output a match or literal. */ +static inline void +lzx_write_item(struct lzx_output_bitstream *os, struct lzx_item item, + unsigned ones_if_aligned, const struct lzx_codes *codes) +{ + u64 data = item.data; + unsigned main_symbol; + unsigned len_symbol; + unsigned num_extra_bits; + u32 extra_bits; + + main_symbol = data & 0x3FF; + + lzx_write_varbits(os, codes->codewords.main[main_symbol], + codes->lens.main[main_symbol], + LZX_MAX_MAIN_CODEWORD_LEN); + + if (main_symbol < LZX_NUM_CHARS) /* Literal? */ + return; + + len_symbol = (data >> 10) & 0xFF; + + if (len_symbol != LZX_LENCODE_NUM_SYMBOLS) { + lzx_write_varbits(os, codes->codewords.len[len_symbol], + codes->lens.len[len_symbol], + LZX_MAX_LEN_CODEWORD_LEN); + } + + num_extra_bits = (data >> 18) & 0x1F; + if (num_extra_bits == 0) /* Small offset or repeat offset match? */ + return; + + extra_bits = data >> 23; + + /*if (block_type == LZX_BLOCKTYPE_ALIGNED && num_extra_bits >= 3) {*/ + if ((num_extra_bits & ones_if_aligned) >= 3) { + + /* Aligned offset blocks: The low 3 bits of the extra offset + * bits are Huffman-encoded using the aligned offset code. The + * remaining bits are output literally. */ + + lzx_write_varbits(os, extra_bits >> 3, num_extra_bits - 3, 14); + + lzx_write_varbits(os, codes->codewords.aligned[extra_bits & 7], + codes->lens.aligned[extra_bits & 7], + LZX_MAX_ALIGNED_CODEWORD_LEN); + } else { + /* Verbatim blocks, or fewer than 3 extra bits: All extra + * offset bits are output literally. */ + lzx_write_varbits(os, extra_bits, num_extra_bits, 17); + } +} + /* - * Writes all compressed matches and literal bytes in an LZX block to the the - * output bitstream. + * Write all matches and literal bytes (which were precomputed) in an LZX + * compressed block to the output bitstream in the final compressed + * representation. * - * @ostream + * @os * The output bitstream. * @block_type - * The type of the block (LZX_BLOCKTYPE_ALIGNED or LZX_BLOCKTYPE_VERBATIM). - * @match_tab - * The array of matches/literals that will be output (length @match_count). - * @match_count - * Number of matches/literals to be output. + * The chosen type of the LZX compressed block (LZX_BLOCKTYPE_ALIGNED or + * LZX_BLOCKTYPE_VERBATIM). + * @items + * The array of matches/literals to output. + * @num_items + * Number of matches/literals to output (length of @items). * @codes - * Pointer to a structure that contains the codewords for the main, length, - * and aligned offset Huffman codes. + * The main, length, and aligned offset Huffman codes for the current + * LZX compressed block. */ static void -lzx_write_matches_and_literals(struct output_bitstream *ostream, - int block_type, - const struct lzx_match match_tab[], - unsigned match_count, - const struct lzx_codes *codes) -{ - for (unsigned i = 0; i < match_count; i++) { - struct lzx_match match = match_tab[i]; - - /* High bit of the match indicates whether the match is an - * actual match (1) or a literal uncompressed byte (0) */ - if (match.data & 0x80000000) { - /* match */ - lzx_write_match(ostream, block_type, - match, codes); - } else { - /* literal byte */ - bitstream_put_bits(ostream, - codes->codewords.main[match.data], - codes->lens.main[match.data]); - } - } -} - -static void -lzx_assert_codes_valid(const struct lzx_codes * codes, unsigned num_main_syms) +lzx_write_items(struct lzx_output_bitstream *os, int block_type, + const struct lzx_item items[], u32 num_items, + const struct lzx_codes *codes) { -#ifdef ENABLE_LZX_DEBUG - unsigned i; - - for (i = 0; i < num_main_syms; i++) - LZX_ASSERT(codes->lens.main[i] <= LZX_MAX_MAIN_CODEWORD_LEN); - - for (i = 0; i < LZX_LENCODE_NUM_SYMBOLS; i++) - LZX_ASSERT(codes->lens.len[i] <= LZX_MAX_LEN_CODEWORD_LEN); + unsigned ones_if_aligned = 0U - (block_type == LZX_BLOCKTYPE_ALIGNED); - for (i = 0; i < LZX_ALIGNEDCODE_NUM_SYMBOLS; i++) - LZX_ASSERT(codes->lens.aligned[i] <= LZX_MAX_ALIGNED_CODEWORD_LEN); - - const unsigned tablebits = 10; - u16 decode_table[(1 << tablebits) + - (2 * max(num_main_syms, LZX_LENCODE_NUM_SYMBOLS))] - _aligned_attribute(DECODE_TABLE_ALIGNMENT); - LZX_ASSERT(0 == make_huffman_decode_table(decode_table, - num_main_syms, - min(tablebits, LZX_MAINCODE_TABLEBITS), - codes->lens.main, - LZX_MAX_MAIN_CODEWORD_LEN)); - LZX_ASSERT(0 == make_huffman_decode_table(decode_table, - LZX_LENCODE_NUM_SYMBOLS, - min(tablebits, LZX_LENCODE_TABLEBITS), - codes->lens.len, - LZX_MAX_LEN_CODEWORD_LEN)); - LZX_ASSERT(0 == make_huffman_decode_table(decode_table, - LZX_ALIGNEDCODE_NUM_SYMBOLS, - min(tablebits, LZX_ALIGNEDCODE_TABLEBITS), - codes->lens.aligned, - LZX_MAX_ALIGNED_CODEWORD_LEN)); -#endif /* ENABLE_LZX_DEBUG */ + for (u32 i = 0; i < num_items; i++) + lzx_write_item(os, items[i], ones_if_aligned, codes); } -/* Write an LZX aligned offset or verbatim block to the output. */ +/* Write an LZX aligned offset or verbatim block to the output bitstream. */ static void lzx_write_compressed_block(int block_type, - unsigned block_size, - unsigned max_window_size, + u32 block_size, + unsigned window_order, unsigned num_main_syms, - struct lzx_match * chosen_matches, - unsigned num_chosen_matches, + struct lzx_item * chosen_items, + u32 num_chosen_items, const struct lzx_codes * codes, - const struct lzx_codes * prev_codes, - struct output_bitstream * ostream) + const struct lzx_lens * prev_lens, + struct lzx_output_bitstream * os) { - unsigned i; - LZX_ASSERT(block_type == LZX_BLOCKTYPE_ALIGNED || block_type == LZX_BLOCKTYPE_VERBATIM); - lzx_assert_codes_valid(codes, num_main_syms); /* The first three bits indicate the type of block and are one of the * LZX_BLOCKTYPE_* constants. */ - bitstream_put_bits(ostream, block_type, 3); + lzx_write_bits(os, block_type, 3); /* Output the block size. * @@ -980,1594 +719,1616 @@ lzx_write_compressed_block(int block_type, * because WIMs created with chunk size greater than 32768 can seemingly * only be opened by wimlib anyway. */ if (block_size == LZX_DEFAULT_BLOCK_SIZE) { - bitstream_put_bits(ostream, 1, 1); + lzx_write_bits(os, 1, 1); } else { - bitstream_put_bits(ostream, 0, 1); + lzx_write_bits(os, 0, 1); - if (max_window_size >= 65536) - bitstream_put_bits(ostream, block_size >> 16, 8); + if (window_order >= 16) + lzx_write_bits(os, block_size >> 16, 8); - bitstream_put_bits(ostream, block_size, 16); + lzx_write_bits(os, block_size & 0xFFFF, 16); } - /* Write out lengths of the main code. Note that the LZX specification - * incorrectly states that the aligned offset code comes after the - * length code, but in fact it is the very first code to be written - * (before the main code). */ - if (block_type == LZX_BLOCKTYPE_ALIGNED) - for (i = 0; i < LZX_ALIGNEDCODE_NUM_SYMBOLS; i++) - bitstream_put_bits(ostream, codes->lens.aligned[i], - LZX_ALIGNEDCODE_ELEMENT_SIZE); - - LZX_DEBUG("Writing main code..."); - - /* Write the precode and lengths for the first LZX_NUM_CHARS symbols in - * the main code, which are the codewords for literal bytes. */ - lzx_write_compressed_code(ostream, - codes->lens.main, - prev_codes->lens.main, - LZX_NUM_CHARS); + /* If it's an aligned offset block, output the aligned offset code. */ + if (block_type == LZX_BLOCKTYPE_ALIGNED) { + for (int i = 0; i < LZX_ALIGNEDCODE_NUM_SYMBOLS; i++) { + lzx_write_bits(os, codes->lens.aligned[i], + LZX_ALIGNEDCODE_ELEMENT_SIZE); + } + } - /* Write the precode and lengths for the rest of the main code, which - * are the codewords for match headers. */ - lzx_write_compressed_code(ostream, - codes->lens.main + LZX_NUM_CHARS, - prev_codes->lens.main + LZX_NUM_CHARS, + /* Output the main code (two parts). */ + lzx_write_compressed_code(os, codes->lens.main, + prev_lens->main, + LZX_NUM_CHARS); + lzx_write_compressed_code(os, codes->lens.main + LZX_NUM_CHARS, + prev_lens->main + LZX_NUM_CHARS, num_main_syms - LZX_NUM_CHARS); - LZX_DEBUG("Writing length code..."); - - /* Write the precode and lengths for the length code. */ - lzx_write_compressed_code(ostream, - codes->lens.len, - prev_codes->lens.len, + /* Output the length code. */ + lzx_write_compressed_code(os, codes->lens.len, + prev_lens->len, LZX_LENCODE_NUM_SYMBOLS); - LZX_DEBUG("Writing matches and literals..."); - - /* Write the actual matches and literals. */ - lzx_write_matches_and_literals(ostream, block_type, - chosen_matches, num_chosen_matches, - codes); - - LZX_DEBUG("Done writing block."); + /* Output the compressed matches and literals. */ + lzx_write_items(os, block_type, chosen_items, num_chosen_items, codes); } -/* Write out the LZX blocks that were computed. */ -static void -lzx_write_all_blocks(struct lzx_compressor *ctx, struct output_bitstream *ostream) +/* Don't allow matches to span the end of an LZX block. */ +static inline unsigned +maybe_truncate_matches(struct lz_match matches[], unsigned num_matches, + struct lzx_compressor *c) { + if (c->match_window_end < c->cur_window_size && num_matches != 0) { + u32 limit = c->match_window_end - c->match_window_pos; - const struct lzx_codes *prev_codes = &ctx->zero_codes; - for (unsigned i = 0; i < ctx->num_blocks; i++) { - const struct lzx_block_spec *spec = &ctx->block_specs[i]; - - LZX_DEBUG("Writing block %u/%u (type=%d, size=%u, num_chosen_matches=%u)...", - i + 1, ctx->num_blocks, - spec->block_type, spec->block_size, - spec->num_chosen_matches); - - lzx_write_compressed_block(spec->block_type, - spec->block_size, - ctx->max_window_size, - ctx->num_main_syms, - &ctx->chosen_matches[spec->chosen_matches_start_pos], - spec->num_chosen_matches, - &spec->codes, - prev_codes, - ostream); - - prev_codes = &spec->codes; + if (limit >= LZX_MIN_MATCH_LEN) { + + unsigned i = num_matches - 1; + do { + if (matches[i].len >= limit) { + matches[i].len = limit; + + /* Truncation might produce multiple + * matches with length 'limit'. Keep at + * most 1. */ + num_matches = i + 1; + } + } while (i--); + } else { + num_matches = 0; + } } + return num_matches; } -/* Constructs an LZX match from a literal byte and updates the main code symbol - * frequencies. */ -static u32 -lzx_tally_literal(u8 lit, struct lzx_freqs *freqs) +static unsigned +lzx_get_matches_fillcache_singleblock(struct lzx_compressor *c, + const struct lz_match **matches_ret) { - freqs->main[lit]++; - return (u32)lit; -} + struct lz_match *cache_ptr; + struct lz_match *matches; + unsigned num_matches; -/* Constructs an LZX match from an offset and a length, and updates the LRU - * queue and the frequency of symbols in the main, length, and aligned offset - * alphabets. The return value is a 32-bit number that provides the match in an - * intermediate representation documented below. */ -static u32 -lzx_tally_match(unsigned match_len, unsigned match_offset, - struct lzx_freqs *freqs, struct lzx_lru_queue *queue) -{ - unsigned position_slot; - unsigned position_footer; - u32 len_header; - unsigned main_symbol; - unsigned len_footer; - unsigned adjusted_match_len; - - LZX_ASSERT(match_len >= LZX_MIN_MATCH_LEN && match_len <= LZX_MAX_MATCH_LEN); - - /* The match offset shall be encoded as a position slot (itself encoded - * as part of the main symbol) and a position footer. */ - position_slot = lzx_get_position_slot(match_offset, queue); - position_footer = (match_offset + LZX_OFFSET_OFFSET) & - ((1U << lzx_get_num_extra_bits(position_slot)) - 1); - - /* The match length shall be encoded as a length header (itself encoded - * as part of the main symbol) and an optional length footer. */ - adjusted_match_len = match_len - LZX_MIN_MATCH_LEN; - if (adjusted_match_len < LZX_NUM_PRIMARY_LENS) { - /* No length footer needed. */ - len_header = adjusted_match_len; + cache_ptr = c->cache_ptr; + matches = cache_ptr + 1; + if (likely(cache_ptr <= c->cache_limit)) { + num_matches = lz_mf_get_matches(c->mf, matches); + cache_ptr->len = num_matches; + c->cache_ptr = matches + num_matches; } else { - /* Length footer needed. It will be encoded using the length - * code. */ - len_header = LZX_NUM_PRIMARY_LENS; - len_footer = adjusted_match_len - LZX_NUM_PRIMARY_LENS; - freqs->len[len_footer]++; + num_matches = 0; } - - /* Account for the main symbol. */ - main_symbol = ((position_slot << 3) | len_header) + LZX_NUM_CHARS; - - freqs->main[main_symbol]++; - - /* In an aligned offset block, 3 bits of the position footer are output - * as an aligned offset symbol. Account for this, although we may - * ultimately decide to output the block as verbatim. */ - - /* The following check is equivalent to: - * - * if (lzx_extra_bits[position_slot] >= 3) - * - * Note that this correctly excludes position slots that correspond to - * recent offsets. */ - if (position_slot >= 8) - freqs->aligned[position_footer & 7]++; - - /* Pack the position slot, position footer, and match length into an - * intermediate representation. See `struct lzx_match' for details. - */ - LZX_ASSERT(LZX_MAX_POSITION_SLOTS <= 64); - LZX_ASSERT(lzx_get_num_extra_bits(LZX_MAX_POSITION_SLOTS - 1) <= 17); - LZX_ASSERT(LZX_MAX_MATCH_LEN - LZX_MIN_MATCH_LEN + 1 <= 256); - - LZX_ASSERT(position_slot <= (1U << (31 - 25)) - 1); - LZX_ASSERT(position_footer <= (1U << (25 - 8)) - 1); - LZX_ASSERT(adjusted_match_len <= (1U << (8 - 0)) - 1); - return 0x80000000 | - (position_slot << 25) | - (position_footer << 8) | - (adjusted_match_len); + c->match_window_pos++; + *matches_ret = matches; + return num_matches; } -struct lzx_record_ctx { - struct lzx_freqs freqs; - struct lzx_lru_queue queue; - struct lzx_match *matches; -}; - -static void -lzx_record_match(unsigned len, unsigned offset, void *_ctx) +static unsigned +lzx_get_matches_fillcache_multiblock(struct lzx_compressor *c, + const struct lz_match **matches_ret) { - struct lzx_record_ctx *ctx = _ctx; + struct lz_match *cache_ptr; + struct lz_match *matches; + unsigned num_matches; - (ctx->matches++)->data = lzx_tally_match(len, offset, &ctx->freqs, &ctx->queue); + cache_ptr = c->cache_ptr; + matches = cache_ptr + 1; + if (likely(cache_ptr <= c->cache_limit)) { + num_matches = lz_mf_get_matches(c->mf, matches); + num_matches = maybe_truncate_matches(matches, num_matches, c); + cache_ptr->len = num_matches; + c->cache_ptr = matches + num_matches; + } else { + num_matches = 0; + } + c->match_window_pos++; + *matches_ret = matches; + return num_matches; } -static void -lzx_record_literal(u8 lit, void *_ctx) +static unsigned +lzx_get_matches_usecache(struct lzx_compressor *c, + const struct lz_match **matches_ret) { - struct lzx_record_ctx *ctx = _ctx; + struct lz_match *cache_ptr; + struct lz_match *matches; + unsigned num_matches; - (ctx->matches++)->data = lzx_tally_literal(lit, &ctx->freqs); + cache_ptr = c->cache_ptr; + matches = cache_ptr + 1; + if (cache_ptr <= c->cache_limit) { + num_matches = cache_ptr->len; + c->cache_ptr = matches + num_matches; + } else { + num_matches = 0; + } + c->match_window_pos++; + *matches_ret = matches; + return num_matches; } -/* Returns the cost, in bits, to output a literal byte using the specified cost - * model. */ static unsigned -lzx_literal_cost(u8 c, const struct lzx_costs * costs) +lzx_get_matches_usecache_nocheck(struct lzx_compressor *c, + const struct lz_match **matches_ret) { - return costs->main[c]; + struct lz_match *cache_ptr; + struct lz_match *matches; + unsigned num_matches; + + cache_ptr = c->cache_ptr; + matches = cache_ptr + 1; + num_matches = cache_ptr->len; + c->cache_ptr = matches + num_matches; + c->match_window_pos++; + *matches_ret = matches; + return num_matches; } -/* Given a (length, offset) pair that could be turned into a valid LZX match as - * well as costs for the codewords in the main, length, and aligned Huffman - * codes, return the approximate number of bits it will take to represent this - * match in the compressed output. Take into account the match offset LRU - * queue and optionally update it. */ static unsigned -lzx_match_cost(unsigned length, unsigned offset, const struct lzx_costs *costs, - struct lzx_lru_queue *queue) +lzx_get_matches_nocache_singleblock(struct lzx_compressor *c, + const struct lz_match **matches_ret) { - unsigned position_slot; - unsigned len_header, main_symbol; - unsigned cost = 0; + struct lz_match *matches; + unsigned num_matches; - position_slot = lzx_get_position_slot(offset, queue); + matches = c->cache_ptr; + num_matches = lz_mf_get_matches(c->mf, matches); + c->match_window_pos++; + *matches_ret = matches; + return num_matches; +} - len_header = min(length - LZX_MIN_MATCH_LEN, LZX_NUM_PRIMARY_LENS); - main_symbol = ((position_slot << 3) | len_header) + LZX_NUM_CHARS; - - /* Account for main symbol. */ - cost += costs->main[main_symbol]; - - /* Account for extra position information. */ - unsigned num_extra_bits = lzx_get_num_extra_bits(position_slot); - if (num_extra_bits >= 3) { - cost += num_extra_bits - 3; - cost += costs->aligned[(offset + LZX_OFFSET_OFFSET) & 7]; - } else { - cost += num_extra_bits; - } - - /* Account for extra length information. */ - if (len_header == LZX_NUM_PRIMARY_LENS) - cost += costs->len[length - LZX_MIN_MATCH_LEN - LZX_NUM_PRIMARY_LENS]; - - return cost; +static unsigned +lzx_get_matches_nocache_multiblock(struct lzx_compressor *c, + const struct lz_match **matches_ret) +{ + struct lz_match *matches; + unsigned num_matches; + matches = c->cache_ptr; + num_matches = lz_mf_get_matches(c->mf, matches); + num_matches = maybe_truncate_matches(matches, num_matches, c); + c->match_window_pos++; + *matches_ret = matches; + return num_matches; } -/* Fast heuristic cost evaluation to use in the inner loop of the match-finder. - * Unlike lzx_match_cost() which does a true cost evaluation, this simply - * prioritize matches based on their offset. */ -static block_cost_t -lzx_match_cost_fast(unsigned offset, const struct lzx_lru_queue *queue) +/* + * Find matches at the next position in the window. + * + * This uses a wrapper function around the underlying match-finder. + * + * Returns the number of matches found and sets *matches_ret to point to the + * matches array. The matches will be sorted by strictly increasing length and + * offset. + */ +static inline unsigned +lzx_get_matches(struct lzx_compressor *c, const struct lz_match **matches_ret) { - /* It seems well worth it to take the time to give priority to recently - * used offsets. */ - for (unsigned i = 0; i < LZX_NUM_RECENT_OFFSETS; i++) - if (offset == queue->R[i]) - return i; - - BUILD_BUG_ON(LZX_MAX_WINDOW_SIZE >= (block_cost_t)~0U); - return offset; + return (*c->get_matches_func)(c, matches_ret); } -/* Set the cost model @ctx->costs from the Huffman codeword lengths specified in - * @lens. - * - * The cost model and codeword lengths are almost the same thing, but the - * Huffman codewords with length 0 correspond to symbols with zero frequency - * that still need to be assigned actual costs. The specific values assigned - * are arbitrary, but they should be fairly high (near the maximum codeword - * length) to take into account the fact that uses of these symbols are expected - * to be rare. */ static void -lzx_set_costs(struct lzx_compressor * ctx, const struct lzx_lens * lens) +lzx_skip_bytes_fillcache(struct lzx_compressor *c, unsigned n) { - unsigned i; - unsigned num_main_syms = ctx->num_main_syms; - - /* Main code */ - for (i = 0; i < num_main_syms; i++) { - ctx->costs.main[i] = lens->main[i]; - if (ctx->costs.main[i] == 0) - ctx->costs.main[i] = ctx->params.alg_params.slow.main_nostat_cost; - } - - /* Length code */ - for (i = 0; i < LZX_LENCODE_NUM_SYMBOLS; i++) { - ctx->costs.len[i] = lens->len[i]; - if (ctx->costs.len[i] == 0) - ctx->costs.len[i] = ctx->params.alg_params.slow.len_nostat_cost; + struct lz_match *cache_ptr; + + cache_ptr = c->cache_ptr; + c->match_window_pos += n; + lz_mf_skip_positions(c->mf, n); + if (cache_ptr <= c->cache_limit) { + do { + cache_ptr->len = 0; + cache_ptr += 1; + } while (--n && cache_ptr <= c->cache_limit); } + c->cache_ptr = cache_ptr; +} - /* Aligned offset code */ - for (i = 0; i < LZX_ALIGNEDCODE_NUM_SYMBOLS; i++) { - ctx->costs.aligned[i] = lens->aligned[i]; - if (ctx->costs.aligned[i] == 0) - ctx->costs.aligned[i] = ctx->params.alg_params.slow.aligned_nostat_cost; +static void +lzx_skip_bytes_usecache(struct lzx_compressor *c, unsigned n) +{ + struct lz_match *cache_ptr; + + cache_ptr = c->cache_ptr; + c->match_window_pos += n; + if (cache_ptr <= c->cache_limit) { + do { + cache_ptr += 1 + cache_ptr->len; + } while (--n && cache_ptr <= c->cache_limit); } + c->cache_ptr = cache_ptr; } -/* Advance the suffix array match-finder to the next position. */ static void -lzx_lz_update_salink(input_idx_t i, - const input_idx_t SA[restrict], - const input_idx_t ISA[restrict], - struct salink link[restrict]) +lzx_skip_bytes_usecache_nocheck(struct lzx_compressor *c, unsigned n) { - /* r = Rank of the suffix at the current position. */ - const input_idx_t r = ISA[i]; - - /* next = rank of LOWEST ranked suffix that is ranked HIGHER than the - * current suffix AND has a LOWER position, or -1 if none exists. */ - const input_idx_t next = link[r].next; - - /* prev = rank of HIGHEST ranked suffix that is ranked LOWER than the - * current suffix AND has a LOWER position, or -1 if none exists. */ - const input_idx_t prev = link[r].prev; - - /* Link the suffix at the current position into the linked list that - * contains all suffixes in the suffix array that are appear at or - * before the current position, sorted by rank. - * - * Save the values of all fields we overwrite so that rollback is - * possible. */ - if (next != (input_idx_t)~0U) { + struct lz_match *cache_ptr; - link[next].prev = r; - link[next].lcpprev = link[r].lcpnext; - } - - if (prev != (input_idx_t)~0U) { + cache_ptr = c->cache_ptr; + c->match_window_pos += n; + do { + cache_ptr += 1 + cache_ptr->len; + } while (--n); + c->cache_ptr = cache_ptr; +} - link[prev].next = r; - link[prev].lcpnext = link[r].lcpprev; - } +static void +lzx_skip_bytes_nocache(struct lzx_compressor *c, unsigned n) +{ + c->match_window_pos += n; + lz_mf_skip_positions(c->mf, n); } /* - * Use the suffix array match-finder to retrieve a list of LZ matches at the - * current position. + * Skip the specified number of positions in the window (don't search for + * matches at them). * - * [in] @i Current position in the window. - * [in] @SA Suffix array for the window. - * [in] @ISA Inverse suffix array for the window. - * [inout] @link Suffix array links used internally by the match-finder. - * [out] @matches The (length, offset) pairs of the resulting matches will - * be written here, sorted in decreasing order by - * length. All returned lengths will be unique. - * [in] @queue Recently used match offsets, used when evaluating the - * cost of matches. - * [in] @min_match_len Minimum match length to return. - * [in] @max_matches_to_consider Maximum number of matches to consider at - * the position. - * [in] @max_matches_to_return Maximum number of matches to return. - * - * The return value is the number of matches found and written to @matches. + * This uses a wrapper function around the underlying match-finder. */ -static unsigned -lzx_lz_get_matches(const input_idx_t i, - const input_idx_t SA[const restrict], - const input_idx_t ISA[const restrict], - struct salink link[const restrict], - struct raw_match matches[const restrict], - const struct lzx_lru_queue * const restrict queue, - const unsigned min_match_len, - const u32 max_matches_to_consider, - const u32 max_matches_to_return) +static inline void +lzx_skip_bytes(struct lzx_compressor *c, unsigned n) { - /* r = Rank of the suffix at the current position. */ - const input_idx_t r = ISA[i]; - - /* Prepare for searching the current position. */ - lzx_lz_update_salink(i, SA, ISA, link); - - /* L = rank of next suffix to the left; - * R = rank of next suffix to the right; - * lenL = length of match between current position and the suffix with rank L; - * lenR = length of match between current position and the suffix with rank R. - * - * This is left and right relative to the rank of the current suffix. - * Since the suffixes in the suffix array are sorted, the longest - * matches are immediately to the left and right (using the linked list - * to ignore all suffixes that occur later in the window). The match - * length decreases the farther left and right we go. We shall keep the - * length on both sides in sync in order to choose the lowest-cost match - * of each length. - */ - input_idx_t L = link[r].prev; - input_idx_t R = link[r].next; - input_idx_t lenL = link[r].lcpprev; - input_idx_t lenR = link[r].lcpnext; - - /* nmatches = number of matches found so far. */ - unsigned nmatches = 0; - - /* best_cost = cost of lowest-cost match found so far. - * - * We keep track of this so that we can ignore shorter matches that do - * not have lower costs than a longer matches already found. - */ - block_cost_t best_cost = INFINITE_BLOCK_COST; + return (*c->skip_bytes_func)(c, n); +} - /* count_remaining = maximum number of possible matches remaining to be - * considered. */ - u32 count_remaining = max_matches_to_consider; +/* Tally, and optionally record, the specified literal byte. */ +static inline void +lzx_declare_literal(struct lzx_compressor *c, unsigned literal, + struct lzx_item **next_chosen_item) +{ + unsigned main_symbol = literal; - /* pending = match currently being considered for a specific length. */ - struct raw_match pending; - block_cost_t pending_cost; + c->freqs.main[main_symbol]++; - while (lenL >= min_match_len || lenR >= min_match_len) - { - pending.len = lenL; - pending_cost = INFINITE_BLOCK_COST; - block_cost_t cost; + if (next_chosen_item) { + *(*next_chosen_item)++ = (struct lzx_item) { + .data = main_symbol, + }; + } +} - /* Extend left. */ - if (lenL >= min_match_len && lenL >= lenR) { - for (;;) { +/* Tally, and optionally record, the specified repeat offset match. */ +static inline void +lzx_declare_repeat_offset_match(struct lzx_compressor *c, + unsigned len, unsigned rep_index, + struct lzx_item **next_chosen_item) +{ + unsigned len_header; + unsigned main_symbol; + unsigned len_symbol; - if (--count_remaining == 0) - goto out_save_pending; + if (len - LZX_MIN_MATCH_LEN < LZX_NUM_PRIMARY_LENS) { + len_header = len - LZX_MIN_MATCH_LEN; + len_symbol = LZX_LENCODE_NUM_SYMBOLS; + } else { + len_header = LZX_NUM_PRIMARY_LENS; + len_symbol = len - LZX_MIN_MATCH_LEN - LZX_NUM_PRIMARY_LENS; + c->freqs.len[len_symbol]++; + } - input_idx_t offset = i - SA[L]; + main_symbol = LZX_NUM_CHARS + ((rep_index << 3) | len_header); - /* Save match if it has smaller cost. */ - cost = lzx_match_cost_fast(offset, queue); - if (cost < pending_cost) { - pending.offset = offset; - pending_cost = cost; - } + c->freqs.main[main_symbol]++; - if (link[L].lcpprev < lenL) { - /* Match length decreased. */ - - lenL = link[L].lcpprev; - - /* Save the pending match unless the - * right side still may have matches of - * this length to be scanned, or if a - * previous (longer) match had lower - * cost. */ - if (pending.len > lenR) { - if (pending_cost < best_cost) { - best_cost = pending_cost; - matches[nmatches++] = pending; - if (nmatches == max_matches_to_return) - return nmatches; - } - pending.len = lenL; - pending_cost = INFINITE_BLOCK_COST; - } - if (lenL < min_match_len || lenL < lenR) - break; - } - L = link[L].prev; - } - } + if (next_chosen_item) { + *(*next_chosen_item)++ = (struct lzx_item) { + .data = (u64)main_symbol | ((u64)len_symbol << 10), + }; + } +} - pending.len = lenR; +/* Tally, and optionally record, the specified explicit offset match. */ +static inline void +lzx_declare_explicit_offset_match(struct lzx_compressor *c, unsigned len, u32 offset, + struct lzx_item **next_chosen_item) +{ + unsigned len_header; + unsigned main_symbol; + unsigned len_symbol; + unsigned offset_slot; + unsigned num_extra_bits; + u32 extra_bits; - /* Extend right. */ - if (lenR >= min_match_len && lenR > lenL) { - for (;;) { + if (len - LZX_MIN_MATCH_LEN < LZX_NUM_PRIMARY_LENS) { + len_header = len - LZX_MIN_MATCH_LEN; + len_symbol = LZX_LENCODE_NUM_SYMBOLS; + } else { + len_header = LZX_NUM_PRIMARY_LENS; + len_symbol = len - LZX_MIN_MATCH_LEN - LZX_NUM_PRIMARY_LENS; + c->freqs.len[len_symbol]++; + } - if (--count_remaining == 0) - goto out_save_pending; + offset_slot = lzx_get_offset_slot_raw(offset + LZX_OFFSET_OFFSET); - input_idx_t offset = i - SA[R]; + main_symbol = LZX_NUM_CHARS + ((offset_slot << 3) | len_header); - /* Save match if it has smaller cost. */ - cost = lzx_match_cost_fast(offset, queue); - if (cost < pending_cost) { - pending.offset = offset; - pending_cost = cost; - } + c->freqs.main[main_symbol]++; - if (link[R].lcpnext < lenR) { - /* Match length decreased. */ + if (offset_slot >= 8) + c->freqs.aligned[(offset + LZX_OFFSET_OFFSET) & 7]++; - lenR = link[R].lcpnext; + if (next_chosen_item) { - /* Save the pending match unless a - * previous (longer) match had lower - * cost. */ - if (pending_cost < best_cost) { - matches[nmatches++] = pending; - best_cost = pending_cost; - if (nmatches == max_matches_to_return) - return nmatches; - } + num_extra_bits = lzx_extra_offset_bits[offset_slot]; - if (lenR < min_match_len || lenR <= lenL) - break; + extra_bits = (offset + LZX_OFFSET_OFFSET) - + lzx_offset_slot_base[offset_slot]; - pending.len = lenR; - pending_cost = INFINITE_BLOCK_COST; - } - R = link[R].next; - } - } + *(*next_chosen_item)++ = (struct lzx_item) { + .data = (u64)main_symbol | + ((u64)len_symbol << 10) | + ((u64)num_extra_bits << 18) | + ((u64)extra_bits << 23), + }; } - goto out; +} -out_save_pending: - if (pending_cost != INFINITE_BLOCK_COST) - matches[nmatches++] = pending; +/* Tally, and optionally record, the specified match or literal. */ +static inline void +lzx_declare_item(struct lzx_compressor *c, u32 mc_item_data, + struct lzx_item **next_chosen_item) +{ + u32 len = mc_item_data & MC_LEN_MASK; + u32 offset_data = mc_item_data >> MC_OFFSET_SHIFT; + + if (len == 1) + lzx_declare_literal(c, offset_data, next_chosen_item); + else if (offset_data < LZX_NUM_RECENT_OFFSETS) + lzx_declare_repeat_offset_match(c, len, offset_data, + next_chosen_item); + else + lzx_declare_explicit_offset_match(c, len, + offset_data - LZX_OFFSET_OFFSET, + next_chosen_item); +} -out: - return nmatches; +static inline void +lzx_record_item_list(struct lzx_compressor *c, + struct lzx_mc_pos_data *cur_optimum_ptr, + struct lzx_item **next_chosen_item) +{ + struct lzx_mc_pos_data *end_optimum_ptr; + u32 saved_item; + u32 item; + + /* The list is currently in reverse order (last item to first item). + * Reverse it. */ + end_optimum_ptr = cur_optimum_ptr; + saved_item = cur_optimum_ptr->mc_item_data; + do { + item = saved_item; + cur_optimum_ptr -= item & MC_LEN_MASK; + saved_item = cur_optimum_ptr->mc_item_data; + cur_optimum_ptr->mc_item_data = item; + } while (cur_optimum_ptr != c->optimum); + + /* Walk the list of items from beginning to end, tallying and recording + * each item. */ + do { + lzx_declare_item(c, cur_optimum_ptr->mc_item_data, next_chosen_item); + cur_optimum_ptr += (cur_optimum_ptr->mc_item_data) & MC_LEN_MASK; + } while (cur_optimum_ptr != end_optimum_ptr); } +static inline void +lzx_tally_item_list(struct lzx_compressor *c, struct lzx_mc_pos_data *cur_optimum_ptr) +{ + /* Since we're just tallying the items, we don't need to reverse the + * list. Processing the items in reverse order is fine. */ + do { + lzx_declare_item(c, cur_optimum_ptr->mc_item_data, NULL); + cur_optimum_ptr -= (cur_optimum_ptr->mc_item_data & MC_LEN_MASK); + } while (cur_optimum_ptr != c->optimum); +} -/* Tell the match-finder to skip the specified number of bytes (@n) in the - * input. */ +/* Tally, and optionally (if next_chosen_item != NULL) record, in order, all + * items in the current list of items found by the match-chooser. */ static void -lzx_lz_skip_bytes(struct lzx_compressor *ctx, unsigned n) +lzx_declare_item_list(struct lzx_compressor *c, struct lzx_mc_pos_data *cur_optimum_ptr, + struct lzx_item **next_chosen_item) { - LZX_ASSERT(n <= ctx->match_window_end - ctx->match_window_pos); - if (ctx->matches_cached) { - ctx->match_window_pos += n; - while (n--) { - ctx->cached_matches_pos += - ctx->cached_matches[ctx->cached_matches_pos].len + 1; - } - } else { - while (n--) { - ctx->cached_matches[ctx->cached_matches_pos++].len = 0; - lzx_lz_update_salink(ctx->match_window_pos++, ctx->SA, - ctx->ISA, ctx->salink); - } - } + if (next_chosen_item) + lzx_record_item_list(c, cur_optimum_ptr, next_chosen_item); + else + lzx_tally_item_list(c, cur_optimum_ptr); } -/* Retrieve a list of matches available at the next position in the input. +/* Set the cost model @c->costs from the Huffman codeword lengths specified in + * @lens. * - * The matches are written to ctx->matches in decreasing order of length, and - * the return value is the number of matches found. */ -static unsigned -lzx_lz_get_matches_caching(struct lzx_compressor *ctx, - const struct lzx_lru_queue *queue, - struct raw_match **matches_ret) + * The cost model and codeword lengths are almost the same thing, but the + * Huffman codewords with length 0 correspond to symbols with zero frequency + * that still need to be assigned actual costs. The specific values assigned + * are arbitrary, but they should be fairly high (near the maximum codeword + * length) to take into account the fact that uses of these symbols are expected + * to be rare. */ +static void +lzx_set_costs(struct lzx_compressor *c, const struct lzx_lens * lens) { - unsigned num_matches; - struct raw_match *matches; + unsigned i; - LZX_ASSERT(ctx->match_window_pos <= ctx->match_window_end); + /* Main code */ + for (i = 0; i < c->num_main_syms; i++) + c->costs.main[i] = lens->main[i] ? lens->main[i] : 15; - matches = &ctx->cached_matches[ctx->cached_matches_pos + 1]; + /* Length code */ + for (i = 0; i < LZX_LENCODE_NUM_SYMBOLS; i++) + c->costs.len[i] = lens->len[i] ? lens->len[i] : 15; - if (ctx->matches_cached) { - num_matches = matches[-1].len; - } else { - unsigned min_match_len = LZX_MIN_MATCH_LEN; - if (!ctx->params.alg_params.slow.use_len2_matches) - min_match_len = max(min_match_len, 3); - const u32 max_search_depth = ctx->params.alg_params.slow.max_search_depth; - const u32 max_matches_per_pos = ctx->params.alg_params.slow.max_matches_per_pos; + /* Aligned offset code */ + for (i = 0; i < LZX_ALIGNEDCODE_NUM_SYMBOLS; i++) + c->costs.aligned[i] = lens->aligned[i] ? lens->aligned[i] : 7; +} - if (unlikely(max_search_depth == 0 || max_matches_per_pos == 0)) - num_matches = 0; - else - num_matches = lzx_lz_get_matches(ctx->match_window_pos, - ctx->SA, - ctx->ISA, - ctx->salink, - matches, - queue, - min_match_len, - max_search_depth, - max_matches_per_pos); - matches[-1].len = num_matches; - } - ctx->cached_matches_pos += num_matches + 1; - *matches_ret = matches; +/* Set default LZX Huffman symbol costs to bootstrap the iterative optimization + * algorithm. */ +static void +lzx_set_default_costs(struct lzx_costs * costs, unsigned num_main_syms) +{ + unsigned i; - /* Cap the length of returned matches to the number of bytes remaining, - * if it is not the whole window. */ - if (ctx->match_window_end < ctx->window_size) { - unsigned maxlen = ctx->match_window_end - ctx->match_window_pos; - for (unsigned i = 0; i < num_matches; i++) - if (matches[i].len > maxlen) - matches[i].len = maxlen; - } -#if 0 - fprintf(stderr, "Pos %u/%u: %u matches\n", - ctx->match_window_pos, ctx->match_window_end, num_matches); - for (unsigned i = 0; i < num_matches; i++) - fprintf(stderr, "\tLen %u Offset %u\n", matches[i].len, matches[i].offset); -#endif + /* Main code (part 1): Literal symbols */ + for (i = 0; i < LZX_NUM_CHARS; i++) + costs->main[i] = 8; -#ifdef ENABLE_LZX_DEBUG - for (unsigned i = 0; i < num_matches; i++) { - LZX_ASSERT(matches[i].len >= LZX_MIN_MATCH_LEN); - LZX_ASSERT(matches[i].len <= LZX_MAX_MATCH_LEN); - LZX_ASSERT(matches[i].len <= ctx->match_window_end - ctx->match_window_pos); - LZX_ASSERT(matches[i].offset > 0); - LZX_ASSERT(matches[i].offset <= ctx->match_window_pos); - LZX_ASSERT(!memcmp(&ctx->window[ctx->match_window_pos], - &ctx->window[ctx->match_window_pos - matches[i].offset], - matches[i].len)); - } -#endif + /* Main code (part 2): Match header symbols */ + for (; i < num_main_syms; i++) + costs->main[i] = 10; - ctx->match_window_pos++; - return num_matches; + /* Length code */ + for (i = 0; i < LZX_LENCODE_NUM_SYMBOLS; i++) + costs->len[i] = 8; + + /* Aligned offset code */ + for (i = 0; i < LZX_ALIGNEDCODE_NUM_SYMBOLS; i++) + costs->aligned[i] = 3; } -/* - * Reverse the linked list of near-optimal matches so that they can be returned - * in forwards order. - * - * Returns the first match in the list. - */ -static struct raw_match -lzx_lz_reverse_near_optimal_match_list(struct lzx_compressor *ctx, - unsigned cur_pos) +/* Return the cost, in bits, to output a literal byte using the specified cost + * model. */ +static inline u32 +lzx_literal_cost(unsigned literal, const struct lzx_costs * costs) { - unsigned prev_link, saved_prev_link; - unsigned prev_match_offset, saved_prev_match_offset; + return costs->main[literal]; +} - ctx->optimum_end_idx = cur_pos; +/* Return the cost, in bits, to output a match of the specified length and + * offset slot using the specified cost model. Does not take into account + * extra offset bits. */ +static inline u32 +lzx_match_cost_raw(unsigned len, unsigned offset_slot, + const struct lzx_costs *costs) +{ + u32 cost; + unsigned len_header; + unsigned main_symbol; - saved_prev_link = ctx->optimum[cur_pos].prev.link; - saved_prev_match_offset = ctx->optimum[cur_pos].prev.match_offset; + if (len - LZX_MIN_MATCH_LEN < LZX_NUM_PRIMARY_LENS) { + len_header = len - LZX_MIN_MATCH_LEN; + cost = 0; + } else { + len_header = LZX_NUM_PRIMARY_LENS; - do { - prev_link = saved_prev_link; - prev_match_offset = saved_prev_match_offset; + /* Account for length symbol. */ + cost = costs->len[len - LZX_MIN_MATCH_LEN - LZX_NUM_PRIMARY_LENS]; + } - saved_prev_link = ctx->optimum[prev_link].prev.link; - saved_prev_match_offset = ctx->optimum[prev_link].prev.match_offset; + /* Account for main symbol. */ + main_symbol = LZX_NUM_CHARS + ((offset_slot << 3) | len_header); + cost += costs->main[main_symbol]; - ctx->optimum[prev_link].next.link = cur_pos; - ctx->optimum[prev_link].next.match_offset = prev_match_offset; + return cost; +} - cur_pos = prev_link; - } while (cur_pos != 0); +/* Equivalent to lzx_match_cost_raw(), but assumes the length is small enough + * that it doesn't require a length symbol. */ +static inline u32 +lzx_match_cost_raw_smalllen(unsigned len, unsigned offset_slot, + const struct lzx_costs *costs) +{ + LZX_ASSERT(len < LZX_MIN_MATCH_LEN + LZX_NUM_PRIMARY_LENS); + return costs->main[LZX_NUM_CHARS + + ((offset_slot << 3) | (len - LZX_MIN_MATCH_LEN))]; +} + +/* + * Consider coding the match at repeat offset index @rep_idx. Consider each + * length from the minimum (2) to the full match length (@rep_len). + */ +static inline void +lzx_consider_repeat_offset_match(struct lzx_compressor *c, + struct lzx_mc_pos_data *cur_optimum_ptr, + unsigned rep_len, unsigned rep_idx) +{ + u32 base_cost = cur_optimum_ptr->cost; + u32 cost; + unsigned len; + +#if 1 /* Optimized version */ + + if (rep_len < LZX_MIN_MATCH_LEN + LZX_NUM_PRIMARY_LENS) { + /* All lengths being considered are small. */ + len = 2; + do { + cost = base_cost + + lzx_match_cost_raw_smalllen(len, rep_idx, &c->costs); + if (cost < (cur_optimum_ptr + len)->cost) { + (cur_optimum_ptr + len)->mc_item_data = + (rep_idx << MC_OFFSET_SHIFT) | len; + (cur_optimum_ptr + len)->cost = cost; + } + } while (++len <= rep_len); + } else { + /* Some lengths being considered are small, and some are big. + * Start with the optimized loop for small lengths, then switch + * to the optimized loop for big lengths. */ + len = 2; + do { + cost = base_cost + + lzx_match_cost_raw_smalllen(len, rep_idx, &c->costs); + if (cost < (cur_optimum_ptr + len)->cost) { + (cur_optimum_ptr + len)->mc_item_data = + (rep_idx << MC_OFFSET_SHIFT) | len; + (cur_optimum_ptr + len)->cost = cost; + } + } while (++len < LZX_MIN_MATCH_LEN + LZX_NUM_PRIMARY_LENS); + + /* The main symbol is now fixed. */ + base_cost += c->costs.main[LZX_NUM_CHARS + + ((rep_idx << 3) | LZX_NUM_PRIMARY_LENS)]; + do { + cost = base_cost + + c->costs.len[len - LZX_MIN_MATCH_LEN - + LZX_NUM_PRIMARY_LENS]; + if (cost < (cur_optimum_ptr + len)->cost) { + (cur_optimum_ptr + len)->mc_item_data = + (rep_idx << MC_OFFSET_SHIFT) | len; + (cur_optimum_ptr + len)->cost = cost; + } + } while (++len <= rep_len); + } - ctx->optimum_cur_idx = ctx->optimum[0].next.link; +#else /* Unoptimized version */ - return (struct raw_match) - { .len = ctx->optimum_cur_idx, - .offset = ctx->optimum[0].next.match_offset, - }; + len = 2; + do { + cost = base_cost + + lzx_match_cost_raw(len, rep_idx, &c->costs); + if (cost < (cur_optimum_ptr + len)->cost) { + (cur_optimum_ptr + len)->mc_item_data = + (rep_idx << MC_OFFSET_SHIFT) | len; + (cur_optimum_ptr + len)->cost = cost; + } + } while (++len <= rep_len); +#endif } /* - * lzx_lz_get_near_optimal_match() - - * - * Choose the optimal match or literal to use at the next position in the input. - * - * Unlike a greedy parser that always takes the longest match, or even a - * parser with one match/literal look-ahead like zlib, the algorithm used here - * may look ahead many matches/literals to determine the optimal match/literal to - * output next. The motivation is that the compression ratio is improved if the - * compressor can do things like use a shorter-than-possible match in order to - * allow a longer match later, and also take into account the Huffman code cost - * model rather than simply assuming that longer is better. - * - * Still, this is not truly an optimal parser because very long matches are - * taken immediately, and the raw match-finder takes some shortcuts. This is - * done to avoid considering many different alternatives that are unlikely to - * be significantly better. - * - * This algorithm is based on that used in 7-Zip's DEFLATE encoder. - * - * Each call to this function does one of two things: - * - * 1. Build a near-optimal sequence of matches/literals, up to some point, that - * will be returned by subsequent calls to this function, then return the - * first one. + * Consider coding each match in @matches as an explicit offset match. * - * OR + * @matches must be sorted by strictly increasing length and strictly + * increasing offset. This is guaranteed by the match-finder. * - * 2. Return the next match/literal previously computed by a call to this - * function; - * - * This function relies on the following state in the compressor context: - * - * ctx->window (read-only: preprocessed data being compressed) - * ctx->cost (read-only: cost model to use) - * ctx->optimum (internal state; leave uninitialized) - * ctx->optimum_cur_idx (must set to 0 before first call) - * ctx->optimum_end_idx (must set to 0 before first call) - * - * Plus any state used by the raw match-finder. - * - * The return value is a (length, offset) pair specifying the match or literal - * chosen. For literals, the length is less than LZX_MIN_MATCH_LEN and the - * offset is meaningless. + * We consider each length from the minimum (2) to the longest + * (matches[num_matches - 1].len). For each length, we consider only + * the smallest offset for which that length is available. Although + * this is not guaranteed to be optimal due to the possibility of a + * larger offset costing less than a smaller offset to code, this is a + * very useful heuristic. */ -static struct raw_match -lzx_lz_get_near_optimal_match(struct lzx_compressor * ctx) +static inline void +lzx_consider_explicit_offset_matches(struct lzx_compressor *c, + struct lzx_mc_pos_data *cur_optimum_ptr, + const struct lz_match matches[], + unsigned num_matches) { - unsigned num_possible_matches; - struct raw_match *possible_matches; - struct raw_match match; - unsigned longest_match_len; - - if (ctx->optimum_cur_idx != ctx->optimum_end_idx) { - /* Case 2: Return the next match/literal already found. */ - match.len = ctx->optimum[ctx->optimum_cur_idx].next.link - - ctx->optimum_cur_idx; - match.offset = ctx->optimum[ctx->optimum_cur_idx].next.match_offset; - - ctx->optimum_cur_idx = ctx->optimum[ctx->optimum_cur_idx].next.link; - return match; + LZX_ASSERT(num_matches > 0); + + unsigned i; + unsigned len; + unsigned offset_slot; + u32 position_cost; + u32 cost; + u32 offset_data; + + +#if 1 /* Optimized version */ + + if (matches[num_matches - 1].offset < LZX_NUM_FAST_OFFSETS) { + + /* + * Offset is small; the offset slot can be looked up directly in + * c->offset_slot_fast. + * + * Additional optimizations: + * + * - Since the offset is small, it falls in the exponential part + * of the offset slot bases and the number of extra offset + * bits can be calculated directly as (offset_slot >> 1) - 1. + * + * - Just consider the number of extra offset bits; don't + * account for the aligned offset code. Usually this has + * almost no effect on the compression ratio. + * + * - Start out in a loop optimized for small lengths. When the + * length becomes high enough that a length symbol will be + * needed, jump into a loop optimized for big lengths. + */ + + LZX_ASSERT(offset_slot <= 37); /* for extra bits formula */ + + len = 2; + i = 0; + do { + offset_slot = c->offset_slot_fast[matches[i].offset]; + position_cost = cur_optimum_ptr->cost + + ((offset_slot >> 1) - 1); + offset_data = matches[i].offset + LZX_OFFSET_OFFSET; + do { + if (len >= LZX_MIN_MATCH_LEN + LZX_NUM_PRIMARY_LENS) + goto biglen; + cost = position_cost + + lzx_match_cost_raw_smalllen(len, offset_slot, + &c->costs); + if (cost < (cur_optimum_ptr + len)->cost) { + (cur_optimum_ptr + len)->cost = cost; + (cur_optimum_ptr + len)->mc_item_data = + (offset_data << MC_OFFSET_SHIFT) | len; + } + } while (++len <= matches[i].len); + } while (++i != num_matches); + + return; + + do { + offset_slot = c->offset_slot_fast[matches[i].offset]; + biglen: + position_cost = cur_optimum_ptr->cost + + ((offset_slot >> 1) - 1) + + c->costs.main[LZX_NUM_CHARS + + ((offset_slot << 3) | + LZX_NUM_PRIMARY_LENS)]; + offset_data = matches[i].offset + LZX_OFFSET_OFFSET; + do { + cost = position_cost + + c->costs.len[len - LZX_MIN_MATCH_LEN - + LZX_NUM_PRIMARY_LENS]; + if (cost < (cur_optimum_ptr + len)->cost) { + (cur_optimum_ptr + len)->cost = cost; + (cur_optimum_ptr + len)->mc_item_data = + (offset_data << MC_OFFSET_SHIFT) | len; + } + } while (++len <= matches[i].len); + } while (++i != num_matches); + } else { + len = 2; + i = 0; + do { + offset_data = matches[i].offset + LZX_OFFSET_OFFSET; + offset_slot = lzx_get_offset_slot_raw(offset_data); + position_cost = cur_optimum_ptr->cost + + lzx_extra_offset_bits[offset_slot]; + do { + cost = position_cost + + lzx_match_cost_raw(len, offset_slot, &c->costs); + if (cost < (cur_optimum_ptr + len)->cost) { + (cur_optimum_ptr + len)->cost = cost; + (cur_optimum_ptr + len)->mc_item_data = + (offset_data << MC_OFFSET_SHIFT) | len; + } + } while (++len <= matches[i].len); + } while (++i != num_matches); } - /* Case 1: Compute a new list of matches/literals to return. */ +#else /* Unoptimized version */ - ctx->optimum_cur_idx = 0; - ctx->optimum_end_idx = 0; + unsigned num_extra_bits; - /* Get matches at this position. */ - num_possible_matches = lzx_lz_get_matches_caching(ctx, &ctx->queue, &possible_matches); + len = 2; + i = 0; + do { + offset_data = matches[i].offset + LZX_OFFSET_OFFSET; + position_cost = cur_optimum_ptr->cost; + offset_slot = lzx_get_offset_slot_raw(offset_data); + num_extra_bits = lzx_extra_offset_bits[offset_slot]; + if (num_extra_bits >= 3) { + position_cost += num_extra_bits - 3; + position_cost += c->costs.aligned[offset_data & 7]; + } else { + position_cost += num_extra_bits; + } + do { + cost = position_cost + + lzx_match_cost_raw(len, offset_slot, &c->costs); + if (cost < (cur_optimum_ptr + len)->cost) { + (cur_optimum_ptr + len)->cost = cost; + (cur_optimum_ptr + len)->mc_item_data = + (offset_data << MC_OFFSET_SHIFT) | len; + } + } while (++len <= matches[i].len); + } while (++i != num_matches); +#endif +} - /* If no matches found, return literal. */ - if (num_possible_matches == 0) - return (struct raw_match){ .len = 0 }; +/* + * Search for repeat offset matches with the current position. + */ +static inline unsigned +lzx_repsearch(const u8 * const strptr, const u32 bytes_remaining, + const struct lzx_lru_queue *queue, unsigned *rep_max_idx_ret) +{ + BUILD_BUG_ON(LZX_NUM_RECENT_OFFSETS != 3); + return lz_repsearch3(strptr, min(bytes_remaining, LZX_MAX_MATCH_LEN), + queue->R, rep_max_idx_ret); +} - /* The matches that were found are sorted in decreasing order by length. - * Get the length of the longest one. */ - longest_match_len = possible_matches[0].len; +/* + * The main near-optimal parsing routine. + * + * Briefly, the algorithm does an approximate minimum-cost path search to find a + * "near-optimal" sequence of matches and literals to output, based on the + * current cost model. The algorithm steps forward, position by position (byte + * by byte), and updates the minimum cost path to reach each later position that + * can be reached using a match or literal from the current position. This is + * essentially Dijkstra's algorithm in disguise: the graph nodes are positions, + * the graph edges are possible matches/literals to code, and the cost of each + * edge is the estimated number of bits that will be required to output the + * corresponding match or literal. But one difference is that we actually + * compute the lowest-cost path in pieces, where each piece is terminated when + * there are no choices to be made. + * + * This function will run this algorithm on the portion of the window from + * &c->cur_window[c->match_window_pos] to &c->cur_window[c->match_window_end]. + * + * On entry, c->queue must be the current state of the match offset LRU queue, + * and c->costs must be the current cost model to use for Huffman symbols. + * + * On exit, c->queue will be the state that the LRU queue would be in if the + * chosen items were to be coded. + * + * If next_chosen_item != NULL, then all items chosen will be recorded (saved in + * the chosen_items array). Otherwise, all items chosen will only be tallied + * (symbol frequencies tallied in c->freqs). + */ +static void +lzx_optim_pass(struct lzx_compressor *c, struct lzx_item **next_chosen_item) +{ + const u8 *block_end; + struct lzx_lru_queue *begin_queue; + const u8 *window_ptr; + struct lzx_mc_pos_data *cur_optimum_ptr; + struct lzx_mc_pos_data *end_optimum_ptr; + const struct lz_match *matches; + unsigned num_matches; + unsigned longest_len; + unsigned rep_max_len; + unsigned rep_max_idx; + unsigned literal; + unsigned len; + u32 cost; + u32 offset_data; + + block_end = &c->cur_window[c->match_window_end]; + begin_queue = &c->queue; +begin: + /* Start building a new list of items, which will correspond to the next + * piece of the overall minimum-cost path. + * + * *begin_queue is the current state of the match offset LRU queue. */ - /* Greedy heuristic: if the longest match that was found is greater - * than the number of fast bytes, return it immediately; don't both - * doing more work. */ - if (longest_match_len > ctx->params.alg_params.slow.num_fast_bytes) { - lzx_lz_skip_bytes(ctx, longest_match_len - 1); - return possible_matches[0]; - } + window_ptr = &c->cur_window[c->match_window_pos]; - /* Calculate the cost to reach the next position by outputting a - * literal. */ - ctx->optimum[0].queue = ctx->queue; - ctx->optimum[1].queue = ctx->optimum[0].queue; - ctx->optimum[1].cost = lzx_literal_cost(ctx->window[ctx->match_window_pos], - &ctx->costs); - ctx->optimum[1].prev.link = 0; - - /* Calculate the cost to reach any position up to and including that - * reached by the longest match, using the shortest (i.e. closest) match - * that reaches each position. */ - BUILD_BUG_ON(LZX_MIN_MATCH_LEN != 2); - for (unsigned len = LZX_MIN_MATCH_LEN, match_idx = num_possible_matches - 1; - len <= longest_match_len; len++) { - - LZX_ASSERT(match_idx < num_possible_matches); - - ctx->optimum[len].queue = ctx->optimum[0].queue; - ctx->optimum[len].prev.link = 0; - ctx->optimum[len].prev.match_offset = possible_matches[match_idx].offset; - ctx->optimum[len].cost = lzx_match_cost(len, - possible_matches[match_idx].offset, - &ctx->costs, - &ctx->optimum[len].queue); - if (len == possible_matches[match_idx].len) - match_idx--; + if (window_ptr == block_end) { + c->queue = *begin_queue; + return; } - unsigned cur_pos = 0; + cur_optimum_ptr = c->optimum; + cur_optimum_ptr->cost = 0; + cur_optimum_ptr->queue = *begin_queue; - /* len_end: greatest index forward at which costs have been calculated - * so far */ - unsigned len_end = longest_match_len; + end_optimum_ptr = cur_optimum_ptr; + /* The following loop runs once for each per byte in the window, except + * in a couple shortcut cases. */ for (;;) { - /* Advance to next position. */ - cur_pos++; - - if (cur_pos == len_end || cur_pos == LZX_OPTIM_ARRAY_SIZE) - return lzx_lz_reverse_near_optimal_match_list(ctx, cur_pos); - - /* retrieve the number of matches available at this position */ - num_possible_matches = lzx_lz_get_matches_caching(ctx, &ctx->optimum[cur_pos].queue, - &possible_matches); - unsigned new_len = 0; + /* Find explicit offset matches with the current position. */ + num_matches = lzx_get_matches(c, &matches); - if (num_possible_matches != 0) { - new_len = possible_matches[0].len; + if (num_matches) { + /* + * Find the longest repeat offset match with the current + * position. + * + * Heuristics: + * + * - Only search for repeat offset matches if the + * match-finder already found at least one match. + * + * - Only consider the longest repeat offset match. It + * seems to be rare for the optimal parse to include a + * repeat offset match that doesn't have the longest + * length (allowing for the possibility that not all + * of that length is actually used). + */ + rep_max_len = lzx_repsearch(window_ptr, + block_end - window_ptr, + &cur_optimum_ptr->queue, + &rep_max_idx); + + if (rep_max_len) { + /* If there's a very long repeat offset match, + * choose it immediately. */ + if (rep_max_len >= c->params.nice_match_length) { + + swap(cur_optimum_ptr->queue.R[0], + cur_optimum_ptr->queue.R[rep_max_idx]); + begin_queue = &cur_optimum_ptr->queue; + + cur_optimum_ptr += rep_max_len; + cur_optimum_ptr->mc_item_data = + (rep_max_idx << MC_OFFSET_SHIFT) | + rep_max_len; + + lzx_skip_bytes(c, rep_max_len - 1); + break; + } - /* Greedy heuristic: if we found a match greater than - * the number of fast bytes, stop immediately. */ - if (new_len > ctx->params.alg_params.slow.num_fast_bytes) { + /* If reaching any positions for the first time, + * initialize their costs to "infinity". */ + while (end_optimum_ptr < cur_optimum_ptr + rep_max_len) + (++end_optimum_ptr)->cost = MC_INFINITE_COST; - /* Build the list of matches to return and get - * the first one. */ - match = lzx_lz_reverse_near_optimal_match_list(ctx, cur_pos); + /* Consider coding a repeat offset match. */ + lzx_consider_repeat_offset_match(c, + cur_optimum_ptr, + rep_max_len, + rep_max_idx); + } - /* Append the long match to the end of the list. */ - ctx->optimum[cur_pos].next.match_offset = - possible_matches[0].offset; - ctx->optimum[cur_pos].next.link = cur_pos + new_len; - ctx->optimum_end_idx = cur_pos + new_len; + longest_len = matches[num_matches - 1].len; + + /* If there's a very long explicit offset match, choose + * it immediately. */ + if (longest_len >= c->params.nice_match_length) { + + cur_optimum_ptr->queue.R[2] = + cur_optimum_ptr->queue.R[1]; + cur_optimum_ptr->queue.R[1] = + cur_optimum_ptr->queue.R[0]; + cur_optimum_ptr->queue.R[0] = + matches[num_matches - 1].offset; + begin_queue = &cur_optimum_ptr->queue; + + offset_data = matches[num_matches - 1].offset + + LZX_OFFSET_OFFSET; + cur_optimum_ptr += longest_len; + cur_optimum_ptr->mc_item_data = + (offset_data << MC_OFFSET_SHIFT) | + longest_len; + + lzx_skip_bytes(c, longest_len - 1); + break; + } - /* Skip over the remaining bytes of the long match. */ - lzx_lz_skip_bytes(ctx, new_len - 1); + /* If reaching any positions for the first time, + * initialize their costs to "infinity". */ + while (end_optimum_ptr < cur_optimum_ptr + longest_len) + (++end_optimum_ptr)->cost = MC_INFINITE_COST; - /* Return first match in the list */ - return match; + /* Consider coding an explicit offset match. */ + lzx_consider_explicit_offset_matches(c, cur_optimum_ptr, + matches, num_matches); + } else { + /* No matches found. The only choice at this position + * is to code a literal. */ + + if (end_optimum_ptr == cur_optimum_ptr) { + #if 1 + /* Optimization for single literals. */ + if (likely(cur_optimum_ptr == c->optimum)) { + lzx_declare_literal(c, *window_ptr++, + next_chosen_item); + if (window_ptr == block_end) { + c->queue = cur_optimum_ptr->queue; + return; + } + continue; + } + #endif + (++end_optimum_ptr)->cost = MC_INFINITE_COST; } } - /* Consider proceeding with a literal byte. */ - block_cost_t cur_cost = ctx->optimum[cur_pos].cost; - block_cost_t cur_plus_literal_cost = cur_cost + - lzx_literal_cost(ctx->window[ctx->match_window_pos - 1], - &ctx->costs); - if (cur_plus_literal_cost < ctx->optimum[cur_pos + 1].cost) { - ctx->optimum[cur_pos + 1].cost = cur_plus_literal_cost; - ctx->optimum[cur_pos + 1].prev.link = cur_pos; - ctx->optimum[cur_pos + 1].queue = ctx->optimum[cur_pos].queue; - } + /* Consider coding a literal. - if (num_possible_matches == 0) - continue; + * To avoid an extra unpredictable brench, actually checking the + * preferability of coding a literal is integrated into the + * queue update code below. */ + literal = *window_ptr++; + cost = cur_optimum_ptr->cost + lzx_literal_cost(literal, &c->costs); + + /* Advance to the next position. */ + cur_optimum_ptr++; + + /* The lowest-cost path to the current position is now known. + * Finalize the recent offsets queue that results from taking + * this lowest-cost path. */ - /* Consider proceeding with a match. */ - - while (len_end < cur_pos + new_len) - ctx->optimum[++len_end].cost = INFINITE_BLOCK_COST; - - for (unsigned len = LZX_MIN_MATCH_LEN, match_idx = num_possible_matches - 1; - len <= new_len; len++) { - LZX_ASSERT(match_idx < num_possible_matches); - struct lzx_lru_queue q = ctx->optimum[cur_pos].queue; - block_cost_t cost = cur_cost + lzx_match_cost(len, - possible_matches[match_idx].offset, - &ctx->costs, - &q); - - if (cost < ctx->optimum[cur_pos + len].cost) { - ctx->optimum[cur_pos + len].cost = cost; - ctx->optimum[cur_pos + len].prev.link = cur_pos; - ctx->optimum[cur_pos + len].prev.match_offset = - possible_matches[match_idx].offset; - ctx->optimum[cur_pos + len].queue = q; + if (cost < cur_optimum_ptr->cost) { + /* Literal: queue remains unchanged. */ + cur_optimum_ptr->cost = cost; + cur_optimum_ptr->mc_item_data = (literal << MC_OFFSET_SHIFT) | 1; + cur_optimum_ptr->queue = (cur_optimum_ptr - 1)->queue; + } else { + /* Match: queue update is needed. */ + len = cur_optimum_ptr->mc_item_data & MC_LEN_MASK; + offset_data = cur_optimum_ptr->mc_item_data >> MC_OFFSET_SHIFT; + if (offset_data >= LZX_NUM_RECENT_OFFSETS) { + /* Explicit offset match: offset is inserted at front */ + cur_optimum_ptr->queue.R[0] = offset_data - LZX_OFFSET_OFFSET; + cur_optimum_ptr->queue.R[1] = (cur_optimum_ptr - len)->queue.R[0]; + cur_optimum_ptr->queue.R[2] = (cur_optimum_ptr - len)->queue.R[1]; + } else { + /* Repeat offset match: offset is swapped to front */ + cur_optimum_ptr->queue = (cur_optimum_ptr - len)->queue; + swap(cur_optimum_ptr->queue.R[0], + cur_optimum_ptr->queue.R[offset_data]); } + } - if (len == possible_matches[match_idx].len) - match_idx--; + /* + * This loop will terminate when either of the following + * conditions is true: + * + * (1) cur_optimum_ptr == end_optimum_ptr + * + * There are no paths that extend beyond the current + * position. In this case, any path to a later position + * must pass through the current position, so we can go + * ahead and choose the list of items that led to this + * position. + * + * (2) cur_optimum_ptr == &c->optimum[LZX_OPTIM_ARRAY_LENGTH] + * + * This bounds the number of times the algorithm can step + * forward before it is guaranteed to start choosing items. + * This limits the memory usage. But + * LZX_OPTIM_ARRAY_LENGTH is high enough that on most + * inputs this limit is never reached. + * + * Note: no check for end-of-block is needed because + * end-of-block will trigger condition (1). + */ + if (cur_optimum_ptr == end_optimum_ptr || + cur_optimum_ptr == &c->optimum[LZX_OPTIM_ARRAY_LENGTH]) + { + begin_queue = &cur_optimum_ptr->queue; + break; } } + + /* Choose the current list of items that constitute the minimum-cost + * path to the current position. */ + lzx_declare_item_list(c, cur_optimum_ptr, next_chosen_item); + goto begin; } -/* - * Set default symbol costs. - */ -static void -lzx_set_default_costs(struct lzx_costs * costs, unsigned num_main_syms) +/* Fast heuristic scoring for lazy parsing: how "good" is this match? */ +static inline unsigned +lzx_explicit_offset_match_score(unsigned len, u32 adjusted_offset) { - unsigned i; + unsigned score = len; - /* Literal symbols */ - for (i = 0; i < LZX_NUM_CHARS; i++) - costs->main[i] = 8; + if (adjusted_offset < 2048) + score++; - /* Match header symbols */ - for (; i < num_main_syms; i++) - costs->main[i] = 10; + if (adjusted_offset < 1024) + score++; - /* Length symbols */ - for (i = 0; i < LZX_LENCODE_NUM_SYMBOLS; i++) - costs->len[i] = 8; - - /* Aligned offset symbols */ - for (i = 0; i < LZX_ALIGNEDCODE_NUM_SYMBOLS; i++) - costs->aligned[i] = 3; + return score; } -/* Given the frequencies of symbols in a compressed block and the corresponding - * Huffman codes, return LZX_BLOCKTYPE_ALIGNED or LZX_BLOCKTYPE_VERBATIM if an - * aligned offset or verbatim block, respectively, will take fewer bits to - * output. */ -static int -lzx_choose_verbatim_or_aligned(const struct lzx_freqs * freqs, - const struct lzx_codes * codes) +static inline unsigned +lzx_repeat_offset_match_score(unsigned len, unsigned slot) { - unsigned aligned_cost = 0; - unsigned verbatim_cost = 0; - - /* Verbatim blocks have a constant 3 bits per position footer. Aligned - * offset blocks have an aligned offset symbol per position footer, plus - * an extra 24 bits to output the lengths necessary to reconstruct the - * aligned offset code itself. */ - for (unsigned i = 0; i < LZX_ALIGNEDCODE_NUM_SYMBOLS; i++) { - verbatim_cost += 3 * freqs->aligned[i]; - aligned_cost += codes->lens.aligned[i] * freqs->aligned[i]; - } - aligned_cost += LZX_ALIGNEDCODE_ELEMENT_SIZE * LZX_ALIGNEDCODE_NUM_SYMBOLS; - if (aligned_cost < verbatim_cost) - return LZX_BLOCKTYPE_ALIGNED; - else - return LZX_BLOCKTYPE_VERBATIM; + return len + 3; } -/* Find a near-optimal sequence of matches/literals with which to output the - * specified LZX block, then set its type to that which has the minimum cost to - * output. */ -static void -lzx_optimize_block(struct lzx_compressor *ctx, struct lzx_block_spec *spec, - unsigned num_passes) +/* Lazy parsing */ +static u32 +lzx_choose_lazy_items_for_block(struct lzx_compressor *c, + u32 block_start_pos, u32 block_size) { - const struct lzx_lru_queue orig_queue = ctx->queue; - struct lzx_freqs freqs; - - unsigned orig_window_pos = spec->window_pos; - unsigned orig_cached_pos = ctx->cached_matches_pos; - - LZX_ASSERT(ctx->match_window_pos == spec->window_pos); + const u8 *window_ptr; + const u8 *block_end; + struct lz_mf *mf; + struct lz_match *matches; + unsigned num_matches; + unsigned cur_len; + u32 cur_offset_data; + unsigned cur_score; + unsigned rep_max_len; + unsigned rep_max_idx; + unsigned rep_score; + unsigned prev_len; + unsigned prev_score; + u32 prev_offset_data; + unsigned skip_len; + struct lzx_item *next_chosen_item; + + window_ptr = &c->cur_window[block_start_pos]; + block_end = window_ptr + block_size; + matches = c->cached_matches; + mf = c->mf; + next_chosen_item = c->chosen_items; + + prev_len = 0; + prev_offset_data = 0; + prev_score = 0; + + while (window_ptr != block_end) { + + /* Find explicit offset matches with the current position. */ + num_matches = lz_mf_get_matches(mf, matches); + window_ptr++; + + if (num_matches == 0 || + (matches[num_matches - 1].len == 3 && + matches[num_matches - 1].offset >= 8192 - LZX_OFFSET_OFFSET && + matches[num_matches - 1].offset != c->queue.R[0] && + matches[num_matches - 1].offset != c->queue.R[1] && + matches[num_matches - 1].offset != c->queue.R[2])) + { + /* No match found, or the only match found was a distant + * length 3 match. Output the previous match if there + * is one; otherwise output a literal. */ - ctx->match_window_end = spec->window_pos + spec->block_size; - spec->chosen_matches_start_pos = spec->window_pos; + no_match_found: - LZX_ASSERT(num_passes >= 1); + if (prev_len) { + skip_len = prev_len - 2; + goto output_prev_match; + } else { + lzx_declare_literal(c, *(window_ptr - 1), + &next_chosen_item); + continue; + } + } - /* The first optimal parsing pass is done using the cost model already - * set in ctx->costs. Each later pass is done using a cost model - * computed from the previous pass. */ - for (unsigned pass = 0; pass < num_passes; pass++) { + /* Find the longest repeat offset match with the current + * position. */ + if (likely(block_end - (window_ptr - 1) >= 2)) { + rep_max_len = lzx_repsearch((window_ptr - 1), + block_end - (window_ptr - 1), + &c->queue, &rep_max_idx); + } else { + rep_max_len = 0; + } - ctx->match_window_pos = orig_window_pos; - ctx->cached_matches_pos = orig_cached_pos; - ctx->queue = orig_queue; - spec->num_chosen_matches = 0; - memset(&freqs, 0, sizeof(freqs)); + cur_len = matches[num_matches - 1].len; + cur_offset_data = matches[num_matches - 1].offset + LZX_OFFSET_OFFSET; + cur_score = lzx_explicit_offset_match_score(cur_len, cur_offset_data); - for (unsigned i = spec->window_pos; i < spec->window_pos + spec->block_size; ) { - struct raw_match raw_match; - struct lzx_match lzx_match; + /* Select the better of the explicit and repeat offset matches. */ + if (rep_max_len >= 3 && + (rep_score = lzx_repeat_offset_match_score(rep_max_len, + rep_max_idx)) >= cur_score) + { + cur_len = rep_max_len; + cur_offset_data = rep_max_idx; + cur_score = rep_score; + } - raw_match = lzx_lz_get_near_optimal_match(ctx); - if (raw_match.len >= LZX_MIN_MATCH_LEN) { - lzx_match.data = lzx_tally_match(raw_match.len, raw_match.offset, - &freqs, &ctx->queue); - i += raw_match.len; - } else { - lzx_match.data = lzx_tally_literal(ctx->window[i], &freqs); - i += 1; - } - ctx->chosen_matches[spec->chosen_matches_start_pos + - spec->num_chosen_matches++] = lzx_match; + if (unlikely(cur_len > block_end - (window_ptr - 1))) { + /* Nearing end of block. */ + cur_len = block_end - (window_ptr - 1); + if (cur_len < 3) + goto no_match_found; } - lzx_make_huffman_codes(&freqs, &spec->codes, - ctx->num_main_syms); - if (pass < num_passes - 1) - lzx_set_costs(ctx, &spec->codes.lens); - ctx->matches_cached = true; - } - spec->block_type = lzx_choose_verbatim_or_aligned(&freqs, &spec->codes); - ctx->matches_cached = false; -} + if (prev_len == 0 || cur_score > prev_score) { + /* No previous match, or the current match is better + * than the previous match. + * + * If there's a previous match, then output a literal in + * its place. + * + * In both cases, if the current match is very long, + * then output it immediately. Otherwise, attempt a + * lazy match by waiting to see if there's a better + * match at the next position. */ -static void -lzx_optimize_blocks(struct lzx_compressor *ctx) -{ - lzx_lru_queue_init(&ctx->queue); - ctx->optimum_cur_idx = 0; - ctx->optimum_end_idx = 0; + if (prev_len) + lzx_declare_literal(c, *(window_ptr - 2), &next_chosen_item); - const unsigned num_passes = ctx->params.alg_params.slow.num_optim_passes; + prev_len = cur_len; + prev_offset_data = cur_offset_data; + prev_score = cur_score; - for (unsigned i = 0; i < ctx->num_blocks; i++) - lzx_optimize_block(ctx, &ctx->block_specs[i], num_passes); -} + if (prev_len >= c->params.nice_match_length) { + skip_len = prev_len - 1; + goto output_prev_match; + } + continue; + } -/* Initialize the suffix array match-finder for the specified input. */ -static void -lzx_lz_init_matchfinder(const u8 T[const restrict], - const input_idx_t n, - input_idx_t SA[const restrict], - input_idx_t ISA[const restrict], - input_idx_t LCP[const restrict], - struct salink link[const restrict], - const unsigned max_match_len) -{ - /* Compute SA (Suffix Array). */ + /* Current match is not better than the previous match, so + * output the previous match. */ - { - /* ISA and link are used as temporary space. */ - BUILD_BUG_ON(LZX_MIN_WINDOW_SIZE * sizeof(ISA[0]) < 256 * sizeof(saidx_t)); - BUILD_BUG_ON(LZX_MIN_WINDOW_SIZE * 2 * sizeof(link[0]) < 256 * 256 * sizeof(saidx_t)); + skip_len = prev_len - 2; - if (sizeof(input_idx_t) == sizeof(saidx_t)) { - divsufsort(T, SA, n, (saidx_t*)ISA, (saidx_t*)link); + output_prev_match: + if (prev_offset_data < LZX_NUM_RECENT_OFFSETS) { + lzx_declare_repeat_offset_match(c, prev_len, + prev_offset_data, + &next_chosen_item); + swap(c->queue.R[0], c->queue.R[prev_offset_data]); } else { - saidx_t sa[n]; - divsufsort(T, sa, n, (saidx_t*)ISA, (saidx_t*)link); - for (input_idx_t i = 0; i < n; i++) - SA[i] = sa[i]; + lzx_declare_explicit_offset_match(c, prev_len, + prev_offset_data - LZX_OFFSET_OFFSET, + &next_chosen_item); + c->queue.R[2] = c->queue.R[1]; + c->queue.R[1] = c->queue.R[0]; + c->queue.R[0] = prev_offset_data - LZX_OFFSET_OFFSET; } + lz_mf_skip_positions(mf, skip_len); + window_ptr += skip_len; + prev_len = 0; } -#ifdef ENABLE_LZX_DEBUG + return next_chosen_item - c->chosen_items; +} - LZX_ASSERT(n > 0); +/* Given the frequencies of symbols in an LZX-compressed block and the + * corresponding Huffman codes, return LZX_BLOCKTYPE_ALIGNED or + * LZX_BLOCKTYPE_VERBATIM if an aligned offset or verbatim block, respectively, + * will take fewer bits to output. */ +static int +lzx_choose_verbatim_or_aligned(const struct lzx_freqs * freqs, + const struct lzx_codes * codes) +{ + u32 aligned_cost = 0; + u32 verbatim_cost = 0; - /* Verify suffix array. */ - { - bool found[n]; - ZERO_ARRAY(found); - for (input_idx_t r = 0; r < n; r++) { - input_idx_t i = SA[r]; - LZX_ASSERT(i < n); - LZX_ASSERT(!found[i]); - found[i] = true; - } + /* A verbatim block requires 3 bits in each place that an aligned symbol + * would be used in an aligned offset block. */ + for (unsigned i = 0; i < LZX_ALIGNEDCODE_NUM_SYMBOLS; i++) { + verbatim_cost += 3 * freqs->aligned[i]; + aligned_cost += codes->lens.aligned[i] * freqs->aligned[i]; } - for (input_idx_t r = 0; r < n - 1; r++) { - - input_idx_t i1 = SA[r]; - input_idx_t i2 = SA[r + 1]; + /* Account for output of the aligned offset code. */ + aligned_cost += LZX_ALIGNEDCODE_ELEMENT_SIZE * LZX_ALIGNEDCODE_NUM_SYMBOLS; - input_idx_t n1 = n - i1; - input_idx_t n2 = n - i2; + if (aligned_cost < verbatim_cost) + return LZX_BLOCKTYPE_ALIGNED; + else + return LZX_BLOCKTYPE_VERBATIM; +} - LZX_ASSERT(memcmp(&T[i1], &T[i2], min(n1, n2)) <= 0); +/* Near-optimal parsing */ +static u32 +lzx_choose_near_optimal_items_for_block(struct lzx_compressor *c, + u32 block_start_pos, u32 block_size) +{ + u32 num_passes_remaining = c->params.num_optim_passes; + struct lzx_lru_queue orig_queue; + struct lzx_item *next_chosen_item; + struct lzx_item **next_chosen_item_ptr; + + /* Choose appropriate match-finder wrapper functions. */ + if (num_passes_remaining > 1) { + if (block_size == c->cur_window_size) + c->get_matches_func = lzx_get_matches_fillcache_singleblock; + else + c->get_matches_func = lzx_get_matches_fillcache_multiblock; + c->skip_bytes_func = lzx_skip_bytes_fillcache; + } else { + if (block_size == c->cur_window_size) + c->get_matches_func = lzx_get_matches_nocache_singleblock; + else + c->get_matches_func = lzx_get_matches_nocache_multiblock; + c->skip_bytes_func = lzx_skip_bytes_nocache; } - LZX_DEBUG("Verified SA (len %u)", n); -#endif /* ENABLE_LZX_DEBUG */ - /* Compute ISA (Inverse Suffix Array) */ - for (input_idx_t r = 0; r < n; r++) - ISA[SA[r]] = r; + /* No matches will extend beyond the end of the block. */ + c->match_window_end = block_start_pos + block_size; - /* Compute LCP (longest common prefix) array. + /* The first optimization pass will use a default cost model. Each + * additional optimization pass will use a cost model computed from the + * previous pass. * - * Algorithm adapted from Kasai et al. 2001: "Linear-Time - * Longest-Common-Prefix Computation in Suffix Arrays and Its - * Applications". */ - { - input_idx_t h = 0; - for (input_idx_t i = 0; i < n; i++) { - input_idx_t r = ISA[i]; - if (r > 0) { - input_idx_t j = SA[r - 1]; - - input_idx_t lim = min(n - i, n - j); - - while (h < lim && T[i + h] == T[j + h]) - h++; - LCP[r] = h; - if (h > 0) - h--; - } + * To improve performance we only generate the array containing the + * matches and literals in intermediate form on the final pass. For + * earlier passes, tallying symbol frequencies is sufficient. */ + lzx_set_default_costs(&c->costs, c->num_main_syms); + + next_chosen_item_ptr = NULL; + orig_queue = c->queue; + do { + /* Reset the match-finder wrapper. */ + c->match_window_pos = block_start_pos; + c->cache_ptr = c->cached_matches; + + if (num_passes_remaining == 1) { + /* Last pass: actually generate the items. */ + next_chosen_item = c->chosen_items; + next_chosen_item_ptr = &next_chosen_item; } - } -#ifdef ENABLE_LZX_DEBUG - /* Verify LCP array. */ - for (input_idx_t r = 0; r < n - 1; r++) { - LZX_ASSERT(ISA[SA[r]] == r); - LZX_ASSERT(ISA[SA[r + 1]] == r + 1); + /* Choose the items. */ + lzx_optim_pass(c, next_chosen_item_ptr); - input_idx_t i1 = SA[r]; - input_idx_t i2 = SA[r + 1]; - input_idx_t lcp = LCP[r + 1]; + if (num_passes_remaining > 1) { + /* This isn't the last pass. */ - input_idx_t n1 = n - i1; - input_idx_t n2 = n - i2; + /* Make the Huffman codes from the symbol frequencies. */ + lzx_make_huffman_codes(&c->freqs, &c->codes[c->codes_index], + c->num_main_syms); - LZX_ASSERT(lcp <= min(n1, n2)); + /* Update symbol costs. */ + lzx_set_costs(c, &c->codes[c->codes_index].lens); - LZX_ASSERT(memcmp(&T[i1], &T[i2], lcp) == 0); - if (lcp < min(n1, n2)) - LZX_ASSERT(T[i1 + lcp] != T[i2 + lcp]); - } -#endif /* ENABLE_LZX_DEBUG */ + /* Reset symbol frequencies. */ + memset(&c->freqs, 0, sizeof(c->freqs)); - /* Compute salink.next and salink.lcpnext. - * - * Algorithm adapted from Crochemore et al. 2009: - * "LPF computation revisited". - * - * Note: we cap lcpnext to the maximum match length so that the - * match-finder need not worry about it later. */ - link[n - 1].next = (input_idx_t)~0U; - link[n - 1].prev = (input_idx_t)~0U; - link[n - 1].lcpnext = 0; - link[n - 1].lcpprev = 0; - for (input_idx_t r = n - 2; r != (input_idx_t)~0U; r--) { - input_idx_t t = r + 1; - input_idx_t l = LCP[t]; - while (t != (input_idx_t)~0 && SA[t] > SA[r]) { - l = min(l, link[t].lcpnext); - t = link[t].next; - } - link[r].next = t; - link[r].lcpnext = min(l, max_match_len); - LZX_ASSERT(t == (input_idx_t)~0U || l <= n - SA[t]); - LZX_ASSERT(l <= n - SA[r]); - LZX_ASSERT(memcmp(&T[SA[r]], &T[SA[t]], l) == 0); - } + /* Reset the match offset LRU queue to what it was at + * the beginning of the block. */ + c->queue = orig_queue; - /* Compute salink.prev and salink.lcpprev. - * - * Algorithm adapted from Crochemore et al. 2009: - * "LPF computation revisited". - * - * Note: we cap lcpprev to the maximum match length so that the - * match-finder need not worry about it later. */ - link[0].prev = (input_idx_t)~0; - link[0].next = (input_idx_t)~0; - link[0].lcpprev = 0; - link[0].lcpnext = 0; - for (input_idx_t r = 1; r < n; r++) { - input_idx_t t = r - 1; - input_idx_t l = LCP[r]; - while (t != (input_idx_t)~0 && SA[t] > SA[r]) { - l = min(l, link[t].lcpprev); - t = link[t].prev; + /* Choose appopriate match-finder wrapper functions. */ + if (c->cache_ptr <= c->cache_limit) { + c->get_matches_func = lzx_get_matches_usecache_nocheck; + c->skip_bytes_func = lzx_skip_bytes_usecache_nocheck; + } else { + c->get_matches_func = lzx_get_matches_usecache; + c->skip_bytes_func = lzx_skip_bytes_usecache; + } } - link[r].prev = t; - link[r].lcpprev = min(l, max_match_len); - LZX_ASSERT(t == (input_idx_t)~0 || l <= n - SA[t]); - LZX_ASSERT(l <= n - SA[r]); - LZX_ASSERT(memcmp(&T[SA[r]], &T[SA[t]], l) == 0); - } -} - -/* Prepare the input window into one or more LZX blocks ready to be output. */ -static void -lzx_prepare_blocks(struct lzx_compressor * ctx) -{ - /* Initialize the match-finder. */ - lzx_lz_init_matchfinder(ctx->window, ctx->window_size, - ctx->SA, ctx->ISA, ctx->LCP, ctx->salink, - LZX_MAX_MATCH_LEN); - ctx->cached_matches_pos = 0; - ctx->matches_cached = false; - ctx->match_window_pos = 0; - - /* Set up a default cost model. */ - lzx_set_default_costs(&ctx->costs, ctx->num_main_syms); - - ctx->num_blocks = DIV_ROUND_UP(ctx->window_size, LZX_DIV_BLOCK_SIZE); - for (unsigned i = 0; i < ctx->num_blocks; i++) { - unsigned pos = LZX_DIV_BLOCK_SIZE * i; - ctx->block_specs[i].window_pos = pos; - ctx->block_specs[i].block_size = min(ctx->window_size - pos, LZX_DIV_BLOCK_SIZE); - } + } while (--num_passes_remaining); - /* Determine sequence of matches/literals to output for each block. */ - lzx_optimize_blocks(ctx); + /* Return the number of items chosen. */ + return next_chosen_item - c->chosen_items; } /* - * This is the fast version of lzx_prepare_blocks(). This version "quickly" - * prepares a single compressed block containing the entire input. See the - * description of the "Fast algorithm" at the beginning of this file for more - * information. - * - * Input --- the preprocessed data: + * Choose the matches/literals with which to output the block of data beginning + * at '&c->cur_window[block_start_pos]' and extending for 'block_size' bytes. * - * ctx->window[] - * ctx->window_size + * The frequences of the Huffman symbols in the block will be tallied in + * 'c->freqs'. * - * Output --- the block specification and the corresponding match/literal data: + * 'c->queue' must specify the state of the queue at the beginning of this block. + * This function will update it to the state of the queue at the end of this + * block. * - * ctx->block_specs[] - * ctx->num_blocks - * ctx->chosen_matches[] + * Returns the number of matches/literals that were chosen and written to + * 'c->chosen_items' in the 'struct lzx_item' intermediate representation. */ -static void -lzx_prepare_block_fast(struct lzx_compressor * ctx) +static u32 +lzx_choose_items_for_block(struct lzx_compressor *c, + u32 block_start_pos, u32 block_size) { - struct lzx_record_ctx record_ctx; - struct lzx_block_spec *spec; - - /* Parameters to hash chain LZ match finder - * (lazy with 1 match lookahead) */ - static const struct lz_params lzx_lz_params = { - /* Although LZX_MIN_MATCH_LEN == 2, length 2 matches typically - * aren't worth choosing when using greedy or lazy parsing. */ - .min_match = 3, - .max_match = LZX_MAX_MATCH_LEN, - .max_offset = 32768, - .good_match = LZX_MAX_MATCH_LEN, - .nice_match = LZX_MAX_MATCH_LEN, - .max_chain_len = LZX_MAX_MATCH_LEN, - .max_lazy_match = LZX_MAX_MATCH_LEN, - .too_far = 4096, - }; - - /* Initialize symbol frequencies and match offset LRU queue. */ - memset(&record_ctx.freqs, 0, sizeof(struct lzx_freqs)); - lzx_lru_queue_init(&record_ctx.queue); - record_ctx.matches = ctx->chosen_matches; - - /* Determine series of matches/literals to output. */ - lz_analyze_block(ctx->window, - ctx->window_size, - lzx_record_match, - lzx_record_literal, - &record_ctx, - &lzx_lz_params, - ctx->prev_tab); - - /* Set up block specification. */ - spec = &ctx->block_specs[0]; - spec->block_type = LZX_BLOCKTYPE_ALIGNED; - spec->window_pos = 0; - spec->block_size = ctx->window_size; - spec->num_chosen_matches = (record_ctx.matches - ctx->chosen_matches); - spec->chosen_matches_start_pos = 0; - lzx_make_huffman_codes(&record_ctx.freqs, &spec->codes, - ctx->num_main_syms); - ctx->num_blocks = 1; + return (*c->params.choose_items_for_block)(c, block_start_pos, block_size); } +/* Initialize c->offset_slot_fast. */ static void -do_call_insn_translation(u32 *call_insn_target, int input_pos, - s32 file_size) +lzx_init_offset_slot_fast(struct lzx_compressor *c) { - s32 abs_offset; - s32 rel_offset; - - rel_offset = le32_to_cpu(*call_insn_target); - if (rel_offset >= -input_pos && rel_offset < file_size) { - if (rel_offset < file_size - input_pos) { - /* "good translation" */ - abs_offset = rel_offset + input_pos; - } else { - /* "compensating translation" */ - abs_offset = rel_offset - file_size; - } - *call_insn_target = cpu_to_le32(abs_offset); - } -} + u8 slot = 0; -/* This is the reverse of undo_call_insn_preprocessing() in lzx-decompress.c. - * See the comment above that function for more information. */ -static void -do_call_insn_preprocessing(u8 data[], int size) -{ - for (int i = 0; i < size - 10; i++) { - if (data[i] == 0xe8) { - do_call_insn_translation((u32*)&data[i + 1], i, - LZX_WIM_MAGIC_FILESIZE); - i += 4; - } - } -} + for (u32 offset = 0; offset < LZX_NUM_FAST_OFFSETS; offset++) { -/* API function documented in wimlib.h */ -WIMLIBAPI unsigned -wimlib_lzx_compress2(const void * const restrict uncompressed_data, - unsigned const uncompressed_len, - void * const restrict compressed_data, - struct wimlib_lzx_context * const restrict lzx_ctx) -{ - struct lzx_compressor *ctx = (struct lzx_compressor*)lzx_ctx; - struct output_bitstream ostream; - input_idx_t compressed_len; + while (offset + LZX_OFFSET_OFFSET >= lzx_offset_slot_base[slot + 1]) + slot++; - if (uncompressed_len < 100) { - LZX_DEBUG("Too small to bother compressing."); - return 0; + c->offset_slot_fast[offset] = slot; } +} - if (uncompressed_len > ctx->max_window_size) { - LZX_DEBUG("Can't compress %u bytes using window of %u bytes!", - uncompressed_len, ctx->max_window_size); - return 0; - } +/* Set internal compression parameters for the specified compression level and + * maximum window size. */ +static void +lzx_build_params(unsigned int compression_level, u32 max_window_size, + struct lzx_compressor_params *lzx_params) +{ + if (compression_level < 25) { - LZX_DEBUG("Attempting to compress %u bytes...", uncompressed_len); + /* Fast compression: Use lazy parsing. */ - /* The input data must be preprocessed. To avoid changing the original - * input, copy it to a temporary buffer. */ - memcpy(ctx->window, uncompressed_data, uncompressed_len); - ctx->window_size = uncompressed_len; + lzx_params->choose_items_for_block = lzx_choose_lazy_items_for_block; + lzx_params->num_optim_passes = 1; - /* This line is unnecessary; it just avoids inconsequential accesses of - * uninitialized memory that would show up in memory-checking tools such - * as valgrind. */ - memset(&ctx->window[ctx->window_size], 0, 12); + /* When lazy parsing, the hash chain match-finding algorithm is + * fastest unless the window is too large. + * + * TODO: something like hash arrays would actually be better + * than binary trees on large windows. */ + if (max_window_size <= 262144) + lzx_params->mf_algo = LZ_MF_HASH_CHAINS; + else + lzx_params->mf_algo = LZ_MF_BINARY_TREES; - LZX_DEBUG("Preprocessing data..."); + /* When lazy parsing, don't bother with length 2 matches. */ + lzx_params->min_match_length = 3; - /* Before doing any actual compression, do the call instruction (0xe8 - * byte) translation on the uncompressed data. */ - do_call_insn_preprocessing(ctx->window, ctx->window_size); + /* Scale nice_match_length and max_search_depth with the + * compression level. */ + lzx_params->nice_match_length = 25 + compression_level * 2; + lzx_params->max_search_depth = 25 + compression_level; + } else { - LZX_DEBUG("Preparing blocks..."); + /* Normal / high compression: Use near-optimal parsing. */ - /* Prepare the compressed data. */ - if (ctx->params.algorithm == WIMLIB_LZX_ALGORITHM_FAST) - lzx_prepare_block_fast(ctx); - else - lzx_prepare_blocks(ctx); + lzx_params->choose_items_for_block = lzx_choose_near_optimal_items_for_block; - LZX_DEBUG("Writing compressed blocks..."); + /* Set a number of optimization passes appropriate for the + * compression level. */ - /* Generate the compressed data. */ - init_output_bitstream(&ostream, compressed_data, ctx->window_size - 1); - lzx_write_all_blocks(ctx, &ostream); + lzx_params->num_optim_passes = 1; - LZX_DEBUG("Flushing bitstream..."); - compressed_len = flush_output_bitstream(&ostream); - if (compressed_len == ~(input_idx_t)0) { - LZX_DEBUG("Data did not compress to less than original length!"); - return 0; - } + if (compression_level >= 40) + lzx_params->num_optim_passes++; - LZX_DEBUG("Done: compressed %u => %u bytes.", - uncompressed_len, compressed_len); - - /* Verify that we really get the same thing back when decompressing. - * Although this could be disabled by default in all cases, it only - * takes around 2-3% of the running time of the slow algorithm to do the - * verification. */ - if (ctx->params.algorithm == WIMLIB_LZX_ALGORITHM_SLOW - #if defined(ENABLE_LZX_DEBUG) || defined(ENABLE_VERIFY_COMPRESSION) - || 1 - #endif - ) - { - /* The decompression buffer can be any temporary space that's no - * longer needed. */ - u8 *buf = (u8*)(ctx->SA ? ctx->SA : ctx->prev_tab); - - if (wimlib_lzx_decompress2(compressed_data, compressed_len, - buf, uncompressed_len, ctx->max_window_size)) - { - ERROR("Failed to decompress data we " - "compressed using LZX algorithm"); - wimlib_assert(0); - return 0; + /* Use more optimization passes for higher compression levels. + * But the more passes there are, the less they help --- so + * don't add them linearly. */ + if (compression_level >= 70) { + lzx_params->num_optim_passes++; + if (compression_level >= 100) + lzx_params->num_optim_passes++; + if (compression_level >= 150) + lzx_params->num_optim_passes++; + if (compression_level >= 200) + lzx_params->num_optim_passes++; + if (compression_level >= 300) + lzx_params->num_optim_passes++; } - if (memcmp(uncompressed_data, buf, uncompressed_len)) { - ERROR("Data we compressed using LZX algorithm " - "didn't decompress to original"); - wimlib_assert(0); - return 0; - } + /* When doing near-optimal parsing, the hash chain match-finding + * algorithm is good if the window size is small and we're only + * doing one optimization pass. Otherwise, the binary tree + * algorithm is the way to go. */ + if (max_window_size <= 32768 && lzx_params->num_optim_passes == 1) + lzx_params->mf_algo = LZ_MF_HASH_CHAINS; + else + lzx_params->mf_algo = LZ_MF_BINARY_TREES; + + /* When doing near-optimal parsing, allow length 2 matches if + * the compression level is sufficiently high. */ + if (compression_level >= 45) + lzx_params->min_match_length = 2; + else + lzx_params->min_match_length = 3; + + /* Scale nice_match_length and max_search_depth with the + * compression level. */ + lzx_params->nice_match_length = min(((u64)compression_level * 32) / 50, + LZX_MAX_MATCH_LEN); + lzx_params->max_search_depth = min(((u64)compression_level * 50) / 50, + LZX_MAX_MATCH_LEN); } - return compressed_len; } -static bool -lzx_params_compatible(const struct wimlib_lzx_params *oldparams, - const struct wimlib_lzx_params *newparams) +/* Given the internal compression parameters and maximum window size, build the + * Lempel-Ziv match-finder parameters. */ +static void +lzx_build_mf_params(const struct lzx_compressor_params *lzx_params, + u32 max_window_size, struct lz_mf_params *mf_params) { - return 0 == memcmp(oldparams, newparams, sizeof(struct wimlib_lzx_params)); + memset(mf_params, 0, sizeof(*mf_params)); + + mf_params->algorithm = lzx_params->mf_algo; + mf_params->max_window_size = max_window_size; + mf_params->min_match_len = lzx_params->min_match_length; + mf_params->max_match_len = LZX_MAX_MATCH_LEN; + mf_params->max_search_depth = lzx_params->max_search_depth; + mf_params->nice_match_len = lzx_params->nice_match_length; } -static struct wimlib_lzx_params lzx_user_default_params; -static struct wimlib_lzx_params *lzx_user_default_params_ptr; +static void +lzx_free_compressor(void *_c); -static bool -lzx_params_valid(const struct wimlib_lzx_params *params) +static u64 +lzx_get_needed_memory(size_t max_block_size, unsigned int compression_level) { - /* Validate parameters. */ - if (params->size_of_this != sizeof(struct wimlib_lzx_params)) { - LZX_DEBUG("Invalid parameter structure size!"); - return false; - } + struct lzx_compressor_params params; + u64 size = 0; + unsigned window_order; + u32 max_window_size; - if (params->algorithm != WIMLIB_LZX_ALGORITHM_SLOW && - params->algorithm != WIMLIB_LZX_ALGORITHM_FAST) - { - LZX_DEBUG("Invalid algorithm."); - return false; - } + window_order = lzx_get_window_order(max_block_size); + if (window_order == 0) + return 0; + max_window_size = max_block_size; - if (params->algorithm == WIMLIB_LZX_ALGORITHM_SLOW) { - if (params->alg_params.slow.num_optim_passes < 1) - { - LZX_DEBUG("Invalid number of optimization passes!"); - return false; - } + lzx_build_params(compression_level, max_window_size, ¶ms); - if (params->alg_params.slow.main_nostat_cost < 1 || - params->alg_params.slow.main_nostat_cost > 16) - { - LZX_DEBUG("Invalid main_nostat_cost!"); - return false; - } + size += sizeof(struct lzx_compressor); - if (params->alg_params.slow.len_nostat_cost < 1 || - params->alg_params.slow.len_nostat_cost > 16) - { - LZX_DEBUG("Invalid len_nostat_cost!"); - return false; - } + /* cur_window */ + size += max_window_size; - if (params->alg_params.slow.aligned_nostat_cost < 1 || - params->alg_params.slow.aligned_nostat_cost > 8) - { - LZX_DEBUG("Invalid aligned_nostat_cost!"); - return false; - } - } - return true; -} + /* mf */ + size += lz_mf_get_needed_memory(params.mf_algo, max_window_size); -/* API function documented in wimlib.h */ -WIMLIBAPI int -wimlib_lzx_set_default_params(const struct wimlib_lzx_params * params) -{ - if (params) { - if (!lzx_params_valid(params)) - return WIMLIB_ERR_INVALID_PARAM; - lzx_user_default_params = *params; - lzx_user_default_params_ptr = &lzx_user_default_params; - } else { - lzx_user_default_params_ptr = NULL; - } - return 0; + /* cached_matches */ + if (params.num_optim_passes > 1) + size += LZX_CACHE_LEN * sizeof(struct lz_match); + else + size += LZX_MAX_MATCHES_PER_POS * sizeof(struct lz_match); + return size; } -/* API function documented in wimlib.h */ -WIMLIBAPI int -wimlib_lzx_alloc_context(u32 window_size, - const struct wimlib_lzx_params *params, - struct wimlib_lzx_context **ctx_pp) +static int +lzx_create_compressor(size_t max_block_size, unsigned int compression_level, + void **c_ret) { - - LZX_DEBUG("Allocating LZX context..."); - - if (!lzx_window_size_valid(window_size)) + struct lzx_compressor *c; + struct lzx_compressor_params params; + struct lz_mf_params mf_params; + unsigned window_order; + u32 max_window_size; + + window_order = lzx_get_window_order(max_block_size); + if (window_order == 0) return WIMLIB_ERR_INVALID_PARAM; + max_window_size = max_block_size; - struct lzx_compressor *ctx; - - static const struct wimlib_lzx_params fast_default = { - .size_of_this = sizeof(struct wimlib_lzx_params), - .algorithm = WIMLIB_LZX_ALGORITHM_FAST, - .use_defaults = 0, - .alg_params = { - .fast = { - }, - }, - }; - static const struct wimlib_lzx_params slow_default = { - .size_of_this = sizeof(struct wimlib_lzx_params), - .algorithm = WIMLIB_LZX_ALGORITHM_SLOW, - .use_defaults = 0, - .alg_params = { - .slow = { - .use_len2_matches = 1, - .num_fast_bytes = 32, - .num_optim_passes = 2, - .max_search_depth = 50, - .max_matches_per_pos = 3, - .main_nostat_cost = 15, - .len_nostat_cost = 15, - .aligned_nostat_cost = 7, - }, - }, - }; - - if (params) { - if (!lzx_params_valid(params)) - return WIMLIB_ERR_INVALID_PARAM; - } else { - LZX_DEBUG("Using default algorithm and parameters."); - if (lzx_user_default_params_ptr) - params = lzx_user_default_params_ptr; - else - params = &slow_default; - } - - if (params->use_defaults) { - if (params->algorithm == WIMLIB_LZX_ALGORITHM_SLOW) - params = &slow_default; - else - params = &fast_default; - } - - if (ctx_pp) { - ctx = *(struct lzx_compressor**)ctx_pp; + lzx_build_params(compression_level, max_window_size, ¶ms); + lzx_build_mf_params(¶ms, max_window_size, &mf_params); + if (!lz_mf_params_valid(&mf_params)) + return WIMLIB_ERR_INVALID_PARAM; - if (ctx && - lzx_params_compatible(&ctx->params, params) && - ctx->max_window_size == window_size) - return 0; + c = CALLOC(1, sizeof(struct lzx_compressor)); + if (!c) + goto oom; + + c->params = params; + c->num_main_syms = lzx_get_num_main_syms(window_order); + c->window_order = window_order; + + /* The window is allocated as 16-byte aligned to speed up memcpy() and + * enable lzx_e8_filter() optimization on x86_64. */ + c->cur_window = ALIGNED_MALLOC(max_window_size, 16); + if (!c->cur_window) + goto oom; + + c->mf = lz_mf_alloc(&mf_params); + if (!c->mf) + goto oom; + + if (params.num_optim_passes > 1) { + c->cached_matches = MALLOC(LZX_CACHE_LEN * + sizeof(struct lz_match)); + if (!c->cached_matches) + goto oom; + c->cache_limit = c->cached_matches + LZX_CACHE_LEN - + (LZX_MAX_MATCHES_PER_POS + 1); } else { - LZX_DEBUG("Check parameters only."); - return 0; + c->cached_matches = MALLOC(LZX_MAX_MATCHES_PER_POS * + sizeof(struct lz_match)); + if (!c->cached_matches) + goto oom; } - LZX_DEBUG("Allocating memory."); - - ctx = CALLOC(1, sizeof(struct lzx_compressor)); - if (ctx == NULL) - goto err; - - ctx->num_main_syms = lzx_get_num_main_syms(window_size); - ctx->max_window_size = window_size; - ctx->window = MALLOC(window_size + 12); - if (ctx->window == NULL) - goto err; + lzx_init_offset_slot_fast(c); - if (params->algorithm == WIMLIB_LZX_ALGORITHM_FAST) { - ctx->prev_tab = MALLOC(window_size * sizeof(ctx->prev_tab[0])); - if (ctx->prev_tab == NULL) - goto err; - } - - size_t block_specs_length = DIV_ROUND_UP(window_size, LZX_DIV_BLOCK_SIZE); - ctx->block_specs = MALLOC(block_specs_length * sizeof(ctx->block_specs[0])); - if (ctx->block_specs == NULL) - goto err; - - if (params->algorithm == WIMLIB_LZX_ALGORITHM_SLOW) { - ctx->SA = MALLOC(3U * window_size * sizeof(ctx->SA[0])); - if (ctx->SA == NULL) - goto err; - ctx->ISA = ctx->SA + window_size; - ctx->LCP = ctx->ISA + window_size; - - ctx->salink = MALLOC(window_size * sizeof(ctx->salink[0])); - if (ctx->salink == NULL) - goto err; - } + *c_ret = c; + return 0; - if (params->algorithm == WIMLIB_LZX_ALGORITHM_SLOW) { - ctx->optimum = MALLOC((LZX_OPTIM_ARRAY_SIZE + LZX_MAX_MATCH_LEN) * - sizeof(ctx->optimum[0])); - if (ctx->optimum == NULL) - goto err; - } +oom: + lzx_free_compressor(c); + return WIMLIB_ERR_NOMEM; +} - if (params->algorithm == WIMLIB_LZX_ALGORITHM_SLOW) { - u32 cache_per_pos; +static size_t +lzx_compress(const void *uncompressed_data, size_t uncompressed_size, + void *compressed_data, size_t compressed_size_avail, void *_c) +{ + struct lzx_compressor *c = _c; + struct lzx_output_bitstream os; + u32 num_chosen_items; + const struct lzx_lens *prev_lens; + u32 block_start_pos; + u32 block_size; + int block_type; - cache_per_pos = params->alg_params.slow.max_matches_per_pos; - if (cache_per_pos > LZX_MAX_CACHE_PER_POS) - cache_per_pos = LZX_MAX_CACHE_PER_POS; + /* Don't bother compressing very small inputs. */ + if (uncompressed_size < 100) + return 0; - ctx->cached_matches = MALLOC(window_size * (cache_per_pos + 1) * - sizeof(ctx->cached_matches[0])); - if (ctx->cached_matches == NULL) - goto err; - } + /* The input data must be preprocessed. To avoid changing the original + * input data, copy it to a temporary buffer. */ + memcpy(c->cur_window, uncompressed_data, uncompressed_size); + c->cur_window_size = uncompressed_size; - ctx->chosen_matches = MALLOC(window_size * sizeof(ctx->chosen_matches[0])); - if (ctx->chosen_matches == NULL) - goto err; + /* Preprocess the data. */ + lzx_do_e8_preprocessing(c->cur_window, c->cur_window_size); - memcpy(&ctx->params, params, sizeof(struct wimlib_lzx_params)); - memset(&ctx->zero_codes, 0, sizeof(ctx->zero_codes)); + /* Load the window into the match-finder. */ + lz_mf_load_window(c->mf, c->cur_window, c->cur_window_size); - LZX_DEBUG("Successfully allocated new LZX context."); + /* Initialize the match offset LRU queue. */ + lzx_lru_queue_init(&c->queue); - wimlib_lzx_free_context(*ctx_pp); - *ctx_pp = (struct wimlib_lzx_context*)ctx; - return 0; + /* Initialize the output bitstream. */ + lzx_init_output(&os, compressed_data, compressed_size_avail); -err: - wimlib_lzx_free_context((struct wimlib_lzx_context*)ctx); - LZX_DEBUG("Ran out of memory."); - return WIMLIB_ERR_NOMEM; + /* Compress the data block by block. + * + * TODO: The compression ratio could be slightly improved by performing + * data-dependent block splitting instead of using fixed-size blocks. + * Doing so well is a computationally hard problem, however. */ + block_start_pos = 0; + c->codes_index = 0; + prev_lens = &c->zero_lens; + do { + /* Compute the block size. */ + block_size = min(LZX_DIV_BLOCK_SIZE, + uncompressed_size - block_start_pos); + + /* Reset symbol frequencies. */ + memset(&c->freqs, 0, sizeof(c->freqs)); + + /* Prepare the matches/literals for the block. */ + num_chosen_items = lzx_choose_items_for_block(c, + block_start_pos, + block_size); + + /* Make the Huffman codes from the symbol frequencies. */ + lzx_make_huffman_codes(&c->freqs, &c->codes[c->codes_index], + c->num_main_syms); + + /* Choose the best block type. + * + * Note: we currently don't consider uncompressed blocks. */ + block_type = lzx_choose_verbatim_or_aligned(&c->freqs, + &c->codes[c->codes_index]); + + /* Write the compressed block to the output buffer. */ + lzx_write_compressed_block(block_type, + block_size, + c->window_order, + c->num_main_syms, + c->chosen_items, + num_chosen_items, + &c->codes[c->codes_index], + prev_lens, + &os); + + /* The current codeword lengths become the previous lengths. */ + prev_lens = &c->codes[c->codes_index].lens; + c->codes_index ^= 1; + + block_start_pos += block_size; + + } while (block_start_pos != uncompressed_size); + + return lzx_flush_output(&os); } -/* API function documented in wimlib.h */ -WIMLIBAPI void -wimlib_lzx_free_context(struct wimlib_lzx_context *_ctx) +static void +lzx_free_compressor(void *_c) { - struct lzx_compressor *ctx = (struct lzx_compressor*)_ctx; - - if (ctx) { - FREE(ctx->chosen_matches); - FREE(ctx->cached_matches); - FREE(ctx->optimum); - FREE(ctx->salink); - FREE(ctx->SA); - FREE(ctx->block_specs); - FREE(ctx->prev_tab); - FREE(ctx->window); - FREE(ctx); - } -} + struct lzx_compressor *c = _c; -/* API function documented in wimlib.h */ -WIMLIBAPI unsigned -wimlib_lzx_compress(const void * const restrict uncompressed_data, - unsigned const uncompressed_len, - void * const restrict compressed_data) -{ - int ret; - struct wimlib_lzx_context *ctx = NULL; - unsigned compressed_len; - - ret = wimlib_lzx_alloc_context(32768, NULL, &ctx); - if (ret) { - wimlib_assert(ret != WIMLIB_ERR_INVALID_PARAM); - WARNING("Couldn't allocate LZX compression context: %"TS"", - wimlib_get_error_string(ret)); - return 0; + if (c) { + ALIGNED_FREE(c->cur_window); + lz_mf_free(c->mf); + FREE(c->cached_matches); + FREE(c); } - - compressed_len = wimlib_lzx_compress2(uncompressed_data, - uncompressed_len, - compressed_data, - ctx); - - wimlib_lzx_free_context(ctx); - - return compressed_len; } + +const struct compressor_ops lzx_compressor_ops = { + .get_needed_memory = lzx_get_needed_memory, + .create_compressor = lzx_create_compressor, + .compress = lzx_compress, + .free_compressor = lzx_free_compressor, +};